The Annals of Statistics

2008, Vol. 36, No. 3, 11711220

DOI: 10.1214/009053607000000677
c Institute of Mathematical Statistics, 2008

arXiv:math/0701907v3 [math.ST] 1 Jul 2008

KERNEL METHODS IN MACHINE LEARNING1

lkopf
By Thomas Hofmann, Bernhard Scho
and Alexander J. Smola
Darmstadt University of Technology, Max Planck Institute for Biological
Cybernetics and National ICT Australia
We review machine learning methods employing positive definite
kernels. These methods formulate learning and estimation problems
in a reproducing kernel Hilbert space (RKHS) of functions defined
on the data domain, expanded in terms of a kernel. Working in linear
spaces of function has the benefit of facilitating the construction and
analysis of learning algorithms while at the same time allowing large
classes of functions. The latter include nonlinear functions as well as
functions defined on nonvectorial data.
We cover a wide range of methods, ranging from binary classifiers
to sophisticated methods for estimation with structured data.

1. Introduction. Over the last ten years estimation and learning methods utilizing positive definite kernels have become rather popular, particularly in machine learning. Since these methods have a stronger mathematical
slant than earlier machine learning methods (e.g., neural networks), there
is also significant interest in the statistics and mathematics community for
these methods. The present review aims to summarize the state of the art on
a conceptual level. In doing so, we build on various sources, including Burges
[25], Cristianini and Shawe-Taylor [37], Herbrich [64] and Vapnik [141] and,
in particular, Sch
olkopf and Smola [118], but we also add a fair amount of
more recent material which helps unifying the exposition. We have not had
space to include proofs; they can be found either in the long version of the
present paper (see Hofmann et al. [69]), in the references given or in the
above books.
The main idea of all the described methods can be summarized in one
paragraph. Traditionally, theory and algorithms of machine learning and
Received December 2005; revised February 2007.
Supported in part by grants of the ARC and by the Pascal Network of Excellence.
AMS 2000 subject classifications. Primary 30C40; secondary 68T05.
Key words and phrases. Machine learning, reproducing kernels, support vector machines, graphical models.
1

This is an electronic reprint of the original article published by the

Institute of Mathematical Statistics in The Annals of Statistics,
2008, Vol. 36, No. 3, 11711220. This reprint differs from the original in
pagination and typographic detail.
1


T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

statistics has been very well developed for the linear case. Real world data
analysis problems, on the other hand, often require nonlinear methods to detect the kind of dependencies that allow successful prediction of properties
of interest. By using a positive definite kernel, one can sometimes have the
best of both worlds. The kernel corresponds to a dot product in a (usually
high-dimensional) feature space. In this space, our estimation methods are
linear, but as long as we can formulate everything in terms of kernel evaluations, we never explicitly have to compute in the high-dimensional feature
space.
The paper has three main sections: Section 2 deals with fundamental
properties of kernels, with special emphasis on (conditionally) positive definite kernels and their characterization. We give concrete examples for such
kernels and discuss kernels and reproducing kernel Hilbert spaces in the context of regularization. Section 3 presents various approaches for estimating
dependencies and analyzing data that make use of kernels. We provide an
overview of the problem formulations as well as their solution using convex
programming techniques. Finally, Section 4 examines the use of reproducing kernel Hilbert spaces as a means to define statistical models, the focus
being on structured, multidimensional responses. We also show how such
techniques can be combined with Markov networks as a suitable framework
to model dependencies between response variables.
2. Kernels.
2.1. An introductory example. Suppose we are given empirical data
(1)

(x1 , y1 ), . . . , (xn , yn ) X Y.

Here, the domain X is some nonempty set that the inputs (the predictor
variables) xi are taken from; the yi Y are called targets (the response variable). Here and below, i, j [n], where we use the notation [n] := {1, . . . , n}.
Note that we have not made any assumptions on the domain X other
than it being a set. In order to study the problem of learning, we need
additional structure. In learning, we want to be able to generalize to unseen
data points. In the case of binary pattern recognition, given some new input
x X , we want to predict the corresponding y {1} (more complex output
domains Y will be treated below). Loosely speaking, we want to choose y
such that (x, y) is in some sense similar to the training examples. To this
end, we need similarity measures in X and in {1}. The latter is easier,
as two target values can only be identical or different. For the former, we
require a function
(2)

k : X X R,

(x, x ) 7 k(x, x )

KERNEL METHODS IN MACHINE LEARNING

Fig. 1. A simple geometric classification algorithm: given two classes of points (depicted by o and +), compute their means c+ , c and assign a test input x to the
one whose mean is closer. This can be done by looking at the dot product between x c
[where c = (c+ + c )/2] and w := c+ c , which changes sign as the enclosed angle passes
through /2. Note that the corresponding decision boundary is a hyperplane (the dotted
line) orthogonal to w (from Sch
olkopf and Smola [118]).

satisfying, for all x, x X ,

k(x, x ) = h(x), (x )i,

(3)

where maps into some dot product space H, sometimes called the feature
space. The similarity measure k is usually called a kernel, and is called its
feature map.
The advantage of using such a kernel as a similarity measure is that
it allows us to construct algorithms in dot product spaces. For instance,
consider the following simple classification algorithm, described in Figure 1,
where Y = {1}. The idea P
is to compute the means of the
two classes in
P
the feature space, c+ = n1+ {i:yi =+1} (xi ), and c = n1 {i:yi =1} (xi ),
where n+ and n are the number of examples with positive and negative
target values, respectively. We then assign a new point (x) to the class
whose mean is closer to it. This leads to the prediction rule
y = sgn(h(x), c+ i h(x), c i + b)

(4)

with b = 21 (kc k2 kc+ k2 ). Substituting the expressions for c yields

!

X
X
1
1
h(x), (xi )i
h(x), (xi )i + b ,
(5) y = sgn
{z
} n
{z
}
|
n+ {i:y =+1} |
{i:y =1}
i

where b = 12 ( n12

k(x,xi )

1
{(i,j):yi =yj =1} k(xi , xj ) n2
+

k(x,xi )

{(i,j):yi =yj =+1} k(xi , xj )).

Let us consider one well-known special case of this type of classifier. Assume that the class means have the same distance to the origin (hence,
b = 0), and that k(, x) is a density for all x X . If the two classes are

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

equally likely and were generated from two probability distributions that
are estimated
X
X
1
1
(6) p+ (x) :=
k(x, xi ),
p (x) :=
k(x, xi ),
n+ {i:y =+1}
n {i:y =1}
i

then (5) is the estimated Bayes decision rule, plugging in the estimates p+
and p for the true densities.
The classifier (5) is closely related to the Support Vector Machine (SVM )
that we will discuss below. It is linear in the feature space (4), while in the
input domain, it is represented by a kernel expansion (5). In both cases, the
decision boundary is a hyperplane in the feature space; however, the normal
vectors [for (4), w = c+ c ] are usually rather different.
The normal vector not only characterizes the alignment of the hyperplane,
its length can also be used to construct tests for the equality of the two classgenerating distributions (Borgwardt et al. [22]).
As an aside, note that if we normalize the targets such that yi = yi /|{j : yj =
yi }|, in which case the yi sum to zero, then kwk2 = hK, yy iF , where h, iF
is the Frobenius dot product. If the two classes have equal size, then up to a
scaling factor involving kKk2 and n, this equals the kernel-target alignment
defined by Cristianini et al. [38].
2.2. Positive definite kernels. We have required that a kernel satisfy (3),
that is, correspond to a dot product in some dot product space. In the
present section we show that the class of kernels that can be written in the
form (3) coincides with the class of positive definite kernels. This has farreaching consequences. There are examples of positive definite kernels which
can be evaluated efficiently even though they correspond to dot products in
infinite dimensional dot product spaces. In such cases, substituting k(x, x )
for h(x), (x )i, as we have done in (5), is crucial. In the machine learning
community, this substitution is called the kernel trick.
Definition 1 (Gram matrix).
X , the n n matrix
(7)

Given a kernel k and inputs x1 , . . . , xn

K := (k(xi , xj ))ij

is called the Gram matrix (or kernel matrix) of k with respect to x1 , . . . , xn .

Definition 2 (Positive definite matrix).
Kij satisfying
(8)

A real n n symmetric matrix

ci cj Kij 0

i,j

for all ci R is called positive definite. If equality in (8) only occurs for
c1 = = cn = 0, then we shall call the matrix strictly positive definite.

KERNEL METHODS IN MACHINE LEARNING

Definition 3 (Positive definite kernel). Let X be a nonempty set. A

function k : X X R which for all n N, xi X , i [n] gives rise to a
positive definite Gram matrix is called a positive definite kernel. A function
k : X X R which for all n N and distinct xi X gives rise to a strictly
positive definite Gram matrix is called a strictly positive definite kernel.
Occasionally, we shall refer to positive definite kernels simply as kernels.
Note that, for simplicity, we have restricted ourselves to the case of real
valued kernels. However, with small changes, the below will also hold for the
complex P
valued case.
P
P
Since i,j ci cj h(xi ), (xj )i = h i ci (xi ), j cj (xj )i 0, kernels of the
form (3) are positive definite for any choice of . In particular, if X is already
a dot product space, we may choose to be the identity. Kernels can thus be
regarded as generalized dot products. While they are not generally bilinear,
they share important properties with dot products, such as the Cauchy
Schwarz inequality: If k is a positive definite kernel, and x1 , x2 X , then
(9)

k(x1 , x2 )2 k(x1 , x1 ) k(x2 , x2 ).

2.2.1. Construction of the reproducing kernel Hilbert space. We now define a map from X into the space of functions mapping X into R, denoted
as RX , via
(10)

: X RX

where x 7 k(, x).

Here, (x) = k(, x) denotes the function that assigns the value k(x , x) to
x X .
We next construct a dot product space containing the images of the inputs
under . To this end, we first turn it into a vector space by forming linear
combinations
(11)

f () =

n
X

i k(, xi ).

i=1

Here, n N, i R and xi X are arbitrary.

Next, we define a dot product between f and another function g() =
P n

j=1 j k(, xj ) (with n N, j R and xj X ) as

(12)

hf, gi :=

n X
n
X

i j k(xi , xj ).

i=1 j=1

To see that this is well defined although

it contains the expansion coefficients
P
and points, note that hf, gi = nj=1 j f (xj ). The latter, however, does not
depend on the particular expansion of f . Similarly, for g, note that hf, gi =
Pn
i=1 i g(xi ). This also shows that h, i is bilinear. It is symmetric, as hf, gi =

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

hg, f i. Moreover, it is positive definite, since positive definiteness of k implies

that, for any function f , written as (11), we have
hf, f i =

(13)

n
X

i j k(xi , xj ) 0.

i,j=1

Next, note that given functions f1 , . . . , fp , and coefficients 1 , . . . , p R, we

have
(14)

p
X

i,j=1

i j hfi , fj i =

p
X

i fi ,

i=1

p
X

j=1

j fj

0.

Here, the equality follows from the bilinearity of h, i, and the right-hand
inequality from (13).
By (14), h, i is a positive definite kernel, defined on our vector space of
functions. For the last step in proving that it even is a dot product, we note
that, by (12), for all functions (11),
(15) hk(, x), f i = f (x) and, in particular, hk(, x), k(, x )i = k(x, x ).
By virtue of these properties, k is called a reproducing kernel (Aronszajn
[7]).
Due to (15) and (9), we have
(16)

|f (x)|2 = |hk(, x), f i|2 k(x, x) hf, f i.

By this inequality, hf, f i = 0 implies f = 0, which is the last property that

was left to prove in order to establish that h, i is a dot product.
Skipping some details, we add that one can complete the space of functions (11) in the norm corresponding to the dot product, and thus gets a
Hilbert space H, called a reproducing kernel Hilbert space (RKHS ).
One can define a RKHS as a Hilbert space H of functions on a set X with
the property that, for all x X and f H, the point evaluations f 7 f (x)
are continuous linear functionals [in particular, all point values f (x) are well
defined, which already distinguishes RKHSs from many L2 Hilbert spaces].
From the point evaluation functional, one can then construct the reproducing kernel using the Riesz representation theorem. The MooreAronszajn
theorem (Aronszajn [7]) states that, for every positive definite kernel on
X X , there exists a unique RKHS and vice versa.
There is an analogue of the kernel trick for distances rather than dot
products, that is, dissimilarities rather than similarities. This leads to the
larger class of conditionally positive definite kernels. Those kernels are defined just like positive definite ones, with the one difference being that their
Gram matrices need to satisfy (8) only subject to
(17)

n
X
i=1

ci = 0.

KERNEL METHODS IN MACHINE LEARNING

Interestingly, it turns out that many kernel algorithms, including SVMs and
kernel PCA (see Section 3), can be applied also with this larger class of
kernels, due to their being translation invariant in feature space (Hein et al.
[63] and Sch
olkopf and Smola [118]).
We conclude this section with a note on terminology. In the early years of
kernel machine learning research, it was not the notion of positive definite
kernels that was being used. Instead, researchers considered kernels satisfying the conditions of Mercers theorem (Mercer [99], see, e.g., Cristianini
and Shawe-Taylor [37] and Vapnik [141]). However, while all such kernels do
satisfy (3), the converse is not true. Since (3) is what we are interested in,
positive definite kernels are thus the right class of kernels to consider.
2.2.2. Properties of positive definite kernels. We begin with some closure
properties of the set of positive definite kernels.
Proposition 4. Below, k1 , k2 , . . . are arbitrary positive definite kernels
on X X , where X is a nonempty set:
(i) The set of positive definite kernels is a closed convex cone, that is,
(a) if 1 , 2 0, then 1 k1 + 2 k2 is positive definite; and (b) if k(x, x ) :=
limn kn (x, x ) exists for all x, x , then k is positive definite.
(ii) The pointwise product k1 k2 is positive definite.
(iii) Assume that for i = 1, 2, ki is a positive definite kernel on Xi Xi ,
where Xi is a nonempty set. Then the tensor product k1 k2 and the direct
sum k1 k2 are positive definite kernels on (X1 X2 ) (X1 X2 ).
The proofs can be found in Berg et al. [18].
It is reassuring that sums and products of positive definite kernels are
positive definite. We will now explain that, loosely speaking, there are no
other operations that preserve positive definiteness. To this end, let C denote the set of all functions : R R that map positive definite kernels to
(conditionally) positive definite kernels (readers who are not interested in
the case of conditionally positive definite kernels may ignore the term in
parentheses). We define
C := {|k is a p.d. kernel (k) is a (conditionally) p.d. kernel},
C = {| for any Hilbert space F,
(hx, x iF ) is (conditionally) positive definite},
C = {| for all n N: K is a p.d.
n n matrix (K) is (conditionally) p.d.},
where (K) is the n n matrix with elements (Kij ).


T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

Proposition 5.

C = C = C .

The following proposition follows from a result of FitzGerald et al. [50] for
(conditionally) positive definite matrices; by Proposition 5, it also applies for
(conditionally) positive definite kernels, and for functions of dot products.
We state the latter case.
Proposition 6. Let : R R. Then (hx, x iF ) is positive definite for
any Hilbert space F if and only if is real entire of the form
(18)

(t) =

an tn

n=0

with an 0 for n 0.
Moreover, (hx, x iF ) is conditionally positive definite for any Hilbert
space F if and only if is real entire of the form (18) with an 0 for
n 1.
There are further properties of k that can be read off the coefficients an :
Steinwart [128] showed that if all an are strictly positive, then the kernel of Proposition 6 is universal on every compact subset S of Rd in the
sense that its RKHS is dense in the space of continuous functions on S in
the k k norm. For support vector machines using universal kernels, he
then shows (universal) consistency (Steinwart [129]). Examples of universal kernels are (19) and (20) below.
In Lemma 11 we will show that the a0 term does not affect an SVM.
Hence, we infer that it is actually sufficient for consistency to have an > 0
for n 1.
We conclude the section with an example of a kernel which is positive definite
by Proposition 6. To this end, let X be a dot product space. The power series
expansion of (x) = ex then tells us that
(19)

k(x, x ) = ehx,x i/

is positive definite (Haussler [62]). If we further multiply k with the positive

2
2
definite kernel f (x)f (x ), where f (x) = ekxk /2 and > 0, this leads to
the positive definiteness of the Gaussian kernel
(20)

2 /(2 2 )

k (x, x ) = k(x, x )f (x)f (x ) = ekxx k

KERNEL METHODS IN MACHINE LEARNING

2.2.3. Properties of positive definite functions. We now let X = Rd and

consider positive definite kernels of the form
(21)

k(x, x ) = h(x x ),

in which case h is called a positive definite function. The following characterization is due to Bochner [21]. We state it in the form given by Wendland
[152].
Theorem 7. A continuous function h on Rd is positive definite if and
only if there exists a finite nonnegative Borel measure on Rd such that
(22)

h(x) =

Rd

eihx,i d().

While normally formulated for complex valued functions, the theorem

also holds true for real functions. Note, however, that if we start with an
arbitrary nonnegative Borel measure, its Fourier transform may not be real.
Real-valued positive definite functions are distinguished by the fact that the
corresponding measures are symmetric.
We may normalize h such that h(0) = 1 [hence, by (9), |h(x)| 1], in
which case is a probability measure and h is its characteristic function. For
2
2
instance, if is a normal distribution of the form (2/ 2 )d/2 e kk /2 d,
2
2
then the corresponding positive definite function is the Gaussian ekxk /(2 ) ;
see (20).
Bochners theorem allows us to interpret the similarity measure k(x, x ) =
h(x x ) in the frequency domain. The choice of the measure determines
which frequency components occur in the kernel. Since the solutions of kernel
algorithms will turn out to be finite kernel expansions, the measure will
thus determine which frequencies occur in the estimates, that is, it will
determine their regularization propertiesmore on that in Section 2.3.2
below.
Bochners theorem generalizes earlier work of Mathias, and has itself been
generalized in various ways, that is, by Schoenberg [115]. An important
generalization considers Abelian semigroups (Berg et al. [18]). In that case,
the theorem provides an integral representation of positive definite functions
in terms of the semigroups semicharacters. Further generalizations were
given by Krein, for the cases of positive definite kernels and functions with
a limited number of negative squares. See Stewart [130] for further details
and references.
As above, there are conditions that ensure that the positive definiteness
becomes strict.
Proposition 8 (Wendland [152]). A positive definite function is strictly
positive definite if the carrier of the measure in its representation (22) contains an open subset.


T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

10

This implies that the Gaussian kernel is strictly positive definite.

An important special case of positive definite functions, which includes
the Gaussian, are radial basis functions. These are functions that can be
written as h(x) = g(kxk2 ) for some function g : [0, [ R. They have the
property of being invariant under the Euclidean group.
2.2.4. Examples of kernels. We have already seen several instances of
positive definite kernels, and now intend to complete our selection with a
few more examples. In particular, we discuss polynomial kernels, convolution
kernels, ANOVA expansions and kernels on documents.
Polynomial kernels. From Proposition 4 it is clear that homogeneous polynomial kernels k(x, x ) = hx, x ip are positive definite for p N and x, x Rd .
By direct calculation, we can derive the corresponding feature map (Poggio
[108]):
p

hx, x i =

d
X

[x]j [x ]j

j=1

(23)
=

!p

[x]j1 [x]jp [x ]j1 [x ]jp = hCp (x), Cp (x )i,

j[d]p

where Cp maps x Rd to the vector Cp (x) whose entries are all possible
pth degree ordered products of the entries of x (note that [d] is used as a
shorthand for {1, . . . , d}). The polynomial kernel of degree p thus computes
a dot product in the space spanned by all monomials of degree p in the input
coordinates. Other useful kernels include the inhomogeneous polynomial,
(24)

k(x, x ) = (hx, x i + c)p

where p N and c 0,

which computes all monomials up to degree p.

Spline kernels. It is possible to obtain spline functions as a result of kernel
expansions (Vapnik et al. [144] simply by noting that convolution of an even
number of indicator functions yields a positive kernel function. Denote by
IX the indicator (or characteristic) function
R on the set X, and denote by
the convolution operation, (f g)(x) := Rd f (x )g(x x) dx . Then the
B-spline kernels are given by
(25)

k(x, x ) = B2p+1 (x x )

where p N with Bi+1 := Bi B0 .

Here B0 is the characteristic function on the unit ball in Rd . From the

definition of (25), it is obvious that, for odd m, we may write Bm as the
inner product between functions Bm/2 . Moreover, note that, for even m, Bm
is not a kernel.

KERNEL METHODS IN MACHINE LEARNING

11

Convolutions and structures. Let us now move to kernels defined on structured objects (Haussler [62] and Watkins [151]). Suppose the object x X is
composed of xp Xp , where p [P ] (note that the sets Xp need not be equal).
For instance, consider the string x = AT G and P = 2. It is composed of the
parts x1 = AT and x2 = G, or alternatively, of x1 = A and x2 = T G. Mathematically speaking, the set of allowed decompositions can be thought
of as a relation R(x1 , . . . , xP , x), to be read as x1 , . . . , xP constitute the
composite object x.
Haussler [62] investigated how to define a kernel between composite objects by building on similarity measures that assess their respective parts;
in other words, kernels kp defined on Xp Xp . Define the R-convolution of
k1 , . . . , kP as
(26)

[k1 kP ](x, x ) :=

P
Y

kp (
xp , x
p ),

x
R(x),
x R(x ) p=1

where the sum runs over all possible ways R(x) and R(x ) in which we
can decompose x into x
1 , . . . , x
P and x analogously [here we used the convention that an empty sum equals zero, hence, if either x or x cannot be
decomposed, then (k1 kP )(x, x ) = 0]. If there is only a finite number
of ways, the relation R is called finite. In this case, it can be shown that the
R-convolution is a valid kernel (Haussler [62]).
ANOVA kernels. Specific examples of convolution kernels are Gaussians
and ANOVA kernels (Vapnik [141] and Wahba [148]). To construct an ANOVA
kernel, we consider X = S N for some set S, and kernels k(i) on S S, where
i = 1, . . . , N . For P = 1, . . . , N , the ANOVA kernel of order P is defined as
(27)

kP (x, x ) :=

P
Y

k(ip ) (xip , xip ).

1i1 <<iP N p=1

Note that if P = N , the sum consists only of the term for which (i1 , . . . , iP ) =
(1, . . . , N ), and k equals the tensor product k(1) k(N ) . At the other
extreme, if P = 1, then the products collapse to one factor each, and k equals
the direct sum k(1) k(N ) . For intermediate values of P , we get kernels
that lie in between tensor products and direct sums.
ANOVA kernels typically use some moderate value of P , which specifies
the order of the interactions between attributes xip that we are interested
in. The sum then runs over the numerous terms that take into account
interactions of order P ; fortunately, the computational cost can be reduced
to O(P d) cost by utilizing recurrent procedures for the kernel evaluation.
ANOVA kernels have been shown to work rather well in multi-dimensional
SV regression problems (Stitson et al. [131]).

12

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

Bag of words. One way in which SVMs have been used for text categorization (Joachims [77]) is the bag-of-words representation. This maps a given
text to a sparse vector, where each component corresponds to a word, and
a component is set to one (or some other number) whenever the related
word occurs in the text. Using an efficient sparse representation, the dot
product between two such vectors can be computed quickly. Furthermore,
this dot product is by construction a valid kernel, referred to as a sparse
vector kernel. One of its shortcomings, however, is that it does not take into
account the word ordering of a document. Other sparse vector kernels are
also conceivable, such as one that maps a text to the set of pairs of words
that are in the same sentence (Joachims [77] and Watkins [151]).
n-grams and suffix trees. A more sophisticated way of dealing with string
data was proposed by Haussler [62] and Watkins [151]. The basic idea is
as described above for general structured objects (26): Compare the strings
by means of the substrings they contain. The more substrings two strings
have in common, the more similar they are. The substrings need not always
be contiguous; that said, the further apart the first and last element of a
substring are, the less weight should be given to the similarity. Depending
on the specific choice of a similarity measure, it is possible to define more
or less efficient kernels which compute the dot product in the feature space
spanned by all substrings of documents.
Consider a finite alphabet , the
set of all strings of length n, n , and
S

the set of all finite strings, := n=0 n . The length of a string s is

denoted by |s|, and its elements by s(1) . . . s(|s|); the concatenation of s and
t is written st. Denote by
k(x, x ) =

#(x, s)#(x , s)cs

a string kernel computed from exact matches. Here #(x, s) is the number of
occurrences of s in x and cs 0.
Vishwanathan and Smola [146] provide an algorithm using suffix trees,
which allows one to compute for arbitrary cs the value of the kernel k(x, x )
in O(|x| + |x |) time and memory. Moreover, also f (x) = hw, (x)i can be
computed in O(|x|) time if preprocessing linear in the size of the support
vectors is carried out. These kernels are then applied to function prediction
(according to the gene ontology) of proteins using only their sequence information. Another prominent application of string kernels is in the field of
splice form prediction and gene finding (R
atsch et al. [112]).
For inexact matches of a limited degree, typically up to = 3, and strings
of bounded length, a similar data structure can be built by explicitly generating a dictionary of strings and their neighborhood in terms of a Hamming
distance (Leslie et al. [92]). These kernels are defined by replacing #(x, s)

KERNEL METHODS IN MACHINE LEARNING

13

by a mismatch function #(x, s, ) which reports the number of approximate

occurrences of s in x. By trading off computational complexity with storage
(hence, the restriction to small numbers of mismatches), essentially lineartime algorithms can be designed. Whether a general purpose algorithm exists
which allows for efficient comparisons of strings with mismatches in linear
time is still an open question.
Mismatch kernels. In the general case it is only possible to find algorithms
whose complexity is linear in the lengths of the documents being compared,
and the length of the substrings, that is, O(|x| |x |) or worse. We now
describe such a kernel with a specific choice of weights (Cristianini and
Shawe-Taylor [37] and Watkins [151]).
Let us now form subsequences u of strings. Given an index sequence i :=
(i1 , . . . , i|u| ) with 1 i1 < < i|u| |s|, we define u := s(i) := s(i1 ) . . . s(i|u| ).
We call l(i) := i|u| i1 + 1 the length of the subsequence in s. Note that if i
is not contiguous, then l(i) > |u|.
The feature space built from strings of length n is defined to be Hn :=
n
R( ) . This notation means that the space has one dimension (or coordinate)
for each element of n , labeled by that element (equivalently, we can think
of it as the space of all real-valued functions on n ). We can thus describe
the feature map coordinate-wise for each u n via
[n (s)]u :=

(28)

l(i) .

i:s(i)=u

Here, 0 < 1 is a decay parameter: The larger the length of the subsequence in s, the smaller the respective contribution to [n (s)]u . The sum
runs over all subsequences of s which equal u.
For instance, consider a dimension of H3 spanned (i.e., labeled) by the
string asd. In this case we have [3 (Nasdaq)]asd = 3 , while [3 (lass das)]asd =
25 . In the first string, asd is a contiguous substring. In the second string,
it appears twice as a noncontiguous substring of length 5 in lass das, the
two occurrences are lass das and lass das.
The kernel induced by the map n takes the form
(29)

kn (s, t) =

un

[n (s)]u [n (t)]u =

l(i) l(j) .

un (i,j):s(i)=t(j)=u

The string kernel kn can be computed using dynamic programming; see

Watkins [151].
The above kernels on string, suffix-tree, mismatch and tree kernels have
been used in sequence analysis. This includes applications in document analysis and categorization, spam filtering, function prediction in proteins, annotations of dna sequences for the detection of introns and exons, named
entity tagging of documents and the construction of parse trees.

14

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

Locality improved kernels. It is possible to adjust kernels to the structure

of spatial data. Recall the Gaussian RBF and polynomial kernels. When
applied to an image, it makes no difference whether one uses as x the image
or a version of x where all locations of the pixels have been permuted. This
indicates that function space on X induced by k does not take advantage of
the locality properties of the data.
By taking advantage of the local structure, estimates can be improved.
On biological sequences (Zien et al. [157]) one may assign more weight to the
entries of the sequence close to the location where estimates should occur.
For images, local interactions between image patches need to be considered. One way is to use the pyramidal kernel (DeCoste and Sch
olkopf [44]
and Sch
olkopf [116]). It takes inner products between corresponding image
patches, then raises the latter to some power p1 , and finally raises their sum
to another power p2 . While the overall degree of this kernel is p1 p2 , the first
factor p1 only captures short range interactions.
Tree kernels. We now discuss similarity measures on more structured objects. For trees Collins and Duffy [31] propose a decomposition method which
maps a tree x into its set of subtrees. The kernel between two trees x, x is
then computed by taking a weighted sum of all terms between both trees.
In particular, Collins and Duffy [31] show a quadratic time algorithm, that
is, O(|x| |x |) to compute this expression, where |x| is the number of nodes
of the tree. When restricting the sum to all proper rooted subtrees, it is
possible to reduce the computational cost to O(|x| + |x |) time by means of
a tree to string conversion (Vishwanathan and Smola [146]).
Graph kernels. Graphs pose a twofold challenge: one may both design a
kernel on vertices of them and also a kernel between them. In the former
case, the graph itself becomes the object defining the metric between the
vertices. See G
artner [56] and Kashima et al. [82] for details on the latter.
In the following we discuss kernels on graphs.
Denote by W Rnn the adjacency matrix of a graph with Wij > 0 if an
edge between i, j exists. Moreover, assume for simplicity that the graph is
undirected, that is, W = W . Denote by L = D W the graph Laplacian
= 1 D 1/2 W D 1/2 the normalized graph Laplacian. Here D is
and by L
P
a diagonal matrix with Dii = j Wij denoting the degree of vertex i.
Fiedler [49] showed that, the second largest eigenvector of L approximately decomposes the graph into two parts according to their sign. The
other large eigenvectors partition the graph into correspondingly smaller
portions. L arises
from the fact that for a function f defined on the vertices
P
of the graph i,j (f (i) f (j))2 = 2f Lf .
Finally, Smola and Kondor [125] show that, under mild conditions and
up to rescaling, L is the only quadratic permutation invariant form which
can be obtained as a linear function of W .

KERNEL METHODS IN MACHINE LEARNING

15

Hence, it is reasonable to consider kernel matrices K obtained from L

Smola and Kondor [125] suggest kernels K = r(L) or K = r(L),

(and L).
which have desirable smoothness properties. Here r : [0, ) [0, ) is a
monotonically decreasing function. Popular choices include
(30)

r() = exp()

(31)

diffusion kernel,

r() = ( + )

regularized graph Laplacian,

(32)

r() = ( )p

p-step random walk,

where > 0 is chosen such as to reflect the amount of diffusion in (30), the
degree of regularization in (31) or the weighting of steps within a random
walk (32) respectively. Equation (30) was proposed by Kondor and Lafferty
[87]. In Section 2.3.2 we will discuss the connection between regularization
operators and kernels in Rn . Without going into details, the function r()
describes the smoothness properties on the graph and L plays the role of
the Laplace operator.
Kernels on sets and subspaces. Whenever each observation xi consists of
a set of instances, we may use a range of methods to capture the specific
properties of these sets (for an overview, see Vishwanathan et al. [147]):
Take
the average of the elements of the set in feature space, that is, (xi ) =
1 P
j (xij ). This yields good performance in the area of multi-instance
n
learning.
Jebara and Kondor [75] extend the idea by dealing with distributions
pi (x) such that (xi ) = E[(x)], where x pi (x). They apply it to image
classification with missing pixels.
Alternatively, one can study angles enclosed by subspaces spanned by
the observations. In a nutshell, if U, U denote the orthogonal matrices
spanning the subspaces of x and x respectively, then k(x, x ) = det U U .
Fisher kernels. [74] have designed kernels building on probability density
models p(x|). Denote by
(33)
(34)

U (x) := log p(x|),

I := Ex [U (x)U (x)],

the Fisher scores and the Fisher information matrix respectively. Note that
for maximum likelihood estimators Ex [U (x)] = 0 and, therefore, I is the
covariance of U (x). The Fisher kernel is defined as
(35)

k(x, x ) := U (x)I 1 U (x ) or k(x, x ) := U (x)U (x )

depending on whether we study the normalized or the unnormalized kernel

respectively.


T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

16

In addition to that, it has several attractive theoretical properties: Oliver

et al. [104] show that estimation using the normalized Fisher kernel corresponds to estimation subject to a regularization on the L2 (p(|)) norm.
Moreover, in the context of exponential families (see Section 4.1 for a
more detailed discussion) where p(x|) = exp(h(x), i g()), we have
(36)

k(x, x ) = [(x) g()][(x ) g()]

for the unnormalized Fisher kernel. This means that up to centering by

g() the Fisher kernel is identical to the kernel arising from the inner
product of the sufficient statistics (x). This is not a coincidence. In fact,
in our analysis of nonparametric exponential families we will encounter this
fact several times (cf. Section 4 for further details). Moreover, note that the
centering is immaterial, as can be seen in Lemma 11.
The above overview of kernel design is by no means complete. The reader
is referred to books of Bakir et al. [9], Cristianini and Shawe-Taylor [37],
Herbrich [64], Joachims [77], Sch
olkopf and Smola [118], Sch
olkopf [121] and
Shawe-Taylor and Cristianini [123] for further examples and details.
2.3. Kernel function classes.
2.3.1. The representer theorem. From kernels, we now move to functions
that can be expressed in terms of kernel expansions. The representer theoolkopf and Smola [118]) shows that
rem (Kimeldorf and Wahba [85] and Sch
solutions of a large class of optimization problems can be expressed as kernel
expansions over the sample points. As above, H is the RKHS associated to
the kernel k.
Theorem 9 (Representer theorem). Denote by : [0, ) R a strictly
monotonic increasing function, by X a set, and by c : (X R2 )n R {}
an arbitrary loss function. Then each minimizer f H of the regularized
risk functional
(37)

c((x1 , y1 , f (x1 )), . . . , (xn , yn , f (xn ))) + (kf k2H )

admits a representation of the form

(38)

f (x) =

n
X

i k(xi , x).

i=1

Monotonicity of does not prevent the regularized risk functional (37)

from having multiple local minima. To ensure a global minimum, we would
need to require convexity. If we discard the strictness of the monotonicity,
then it no longer follows that each minimizer of the regularized risk admits

KERNEL METHODS IN MACHINE LEARNING

17

an expansion (38); it still follows, however, that there is always another

solution that is as good, and that does admit the expansion.
The significance of the representer theorem is that although we might be
trying to solve an optimization problem in an infinite-dimensional space H,
containing linear combinations of kernels centered on arbitrary points of X ,
it states that the solution lies in the span of n particular kernelsthose
centered on the training points. We will encounter (38) again further below,
where it is called the Support Vector expansion. For suitable choices of loss
functions, many of the i often equal 0.
Despite the finiteness of the representation in (38), it can often be the
case that the number of terms in the expansion is too large in practice.
This can be problematic in practice, since the time required to evaluate (38)
is proportional to the number of terms. One can reduce this number by
computing a reduced representation which approximates the original one in
the RKHS norm (e.g., Sch
olkopf and Smola [118]).
2.3.2. Regularization properties. The regularizer kf k2H used in Theorem 9,
which is what distinguishes SVMs from many other regularized function estimators (e.g., based on coefficient based L1 regularizers, such as the Lasso
olkopf and Smola
(Tibshirani [135]) or linear programming machines (Sch
[118])), stems from the dot product hf, f ik in the RKHS H associated with a
positive definite kernel. The nature and implications of this regularizer, however, are not obvious and we shall now provide an analysis in the Fourier domain. It turns out that if the kernel is translation invariant, then its Fourier
transform allows us to characterize how the different frequency components
of f contribute to the value of kf k2H . Our exposition will be informal (see
also Poggio and Girosi [109] and Smola et al. [127]), and we will implicitly
assume that all integrals are over Rd and exist, and that the operators are
well defined.
We will rewrite the RKHS dot product as
(39)

hf, gik = hf, gi = h2 f, gi,

where is a positive (and thus symmetric) operator mapping H into a

function space endowed with the usual dot product
(40)

hf, gi =

f (x)g(x) dx.

Rather than (39), we consider the equivalent condition (cf. Section 2.2.1)
(41)

hk(x, ), k(x , )ik = hk(x, ), k(x , )i = h2 k(x, ), k(x , )i.

If k(x, ) is a Green function of 2 , we have

(42)

h2 k(x, ), k(x , )i = hx , k(x , )i = k(x, x ),


T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

18

which by the reproducing property (15) amounts to the desired equality

(41).
For conditionally positive definite kernels, a similar correspondence can
be established, with a regularization operator whose null space is spanned
by a set of functions which are not regularized [in the case (17), which is
sometimes called conditionally positive definite of order 1, these are the
constants].
We now consider the particular case where the kernel can be written
k(x, x ) = h(x x ) with a continuous strictly positive definite function
h L1 (Rd ) (cf. Section 2.2.3). A variation of Bochners theorem, stated by
Wendland [152], then tells us that the measure corresponding to h has a
nonvanishing density with respect to the Lebesgue measure, that is, that
k can be written as
(43)

k(x, x ) =

ihxx ,i

() d =

eihx,i eihx ,i () d.

We would like to rewrite this as hk(x, ), k(x , )i for some linear operator
. It turns out that a multiplication operator in the Fourier domain will do
the job. To this end, recall the d-dimensional Fourier transform, given by
(44)

F [f ]() := (2)d/2

(45)

with the inverse F

d/2

[f ](x) = (2)

Next, compute the Fourier transform of k as

F [k(x, )]() = (2)d/2
(46)

Z Z

f (x)eihx,i dx,
f ()eihx,i d.

(( )eihx, i )eihx , i d eihx ,i dx

= (2)d/2 ()eihx,i .

Hence, we can rewrite (43) as

(47)

k(x, x ) = (2)d

F [k(x, )]()F [k(x , )]()

d.
()

If our regularization operator maps

: f 7 (2)d/2 1/2 F [f ],

(48)
we thus have
(49)

k(x, x ) =

(k(x, ))()(k(x , ))() d,

that is, our desired identity (41) holds true.

As required in (39), we
R can thus interpret the dot product hf, gik in the
RKHS as a dot product (f )()(g)() d. This allows us to understand

KERNEL METHODS IN MACHINE LEARNING

19

regularization properties of k in terms of its (scaled) Fourier transform ().

Small values of () amplify the corresponding frequencies in (48). Penalizing hf, f ik thus amounts to a strong attenuation of the corresponding
frequencies. Hence, small values of () for large kk are desirable, since
high-frequency components of F [f ] correspond to rapid changes in f . It
follows that () describes the filter properties of the corresponding regularization operator . In view of our comments following Theorem 7, we
can translate this insight into probabilistic terms: if the probability measure
R() d describes the desired filter properties, then the natural translation
() d
invariant kernel to use is the characteristic function of the measure.

2.3.3. Remarks and notes. The notion of kernels as dot products in

Hilbert spaces was brought to the field of machine learning by Aizerman
olkopf at al. [119] and Vapnik [141]. Aizerman
et al. [1], Boser at al. [23], Sch
et al. [1] used kernels as a tool in a convergence proof, allowing them to apply the Perceptron convergence theorem to their class of potential function
algorithms. To the best of our knowledge, Boser et al. [23] were the first to
use kernels to construct a nonlinear estimation algorithm, the hard margin
predecessor of the Support Vector Machine, from its linear counterpart, the
generalized portrait (Vapnik [139] and Vapnik and Lerner [145]). While all
these uses were limited to kernels defined on vectorial data, Sch
olkopf [116]
observed that this restriction is unnecessary, and nontrivial kernels on other
data types were proposed by Haussler [62] and Watkins [151]. Sch
olkopf et al.
[119] applied the kernel trick to generalize principal component analysis and
pointed out the (in retrospect obvious) fact that any algorithm which only
uses the data via dot products can be generalized using kernels.
In addition to the above uses of positive definite kernels in machine learning, there has been a parallel, and partly earlier development in the field of
statistics, where such kernels have been used, for instance, for time series
analysis (Parzen [106]), as well as regression estimation and the solution of
inverse problems (Wahba [148]).
In probability theory, positive definite kernels have also been studied in
depth since they arise as covariance kernels of stochastic processes; see,
for example, Lo`eve [93]. This connection is heavily being used in a subset
of the machine learning community interested in prediction with Gaussian
processes (Rasmussen and Williams [111]).
In functional analysis, the problem of Hilbert space representations of
kernels has been studied in great detail; a good reference is Berg at al. [18];
indeed, a large part of the material in the present section is based on that
work. Interestingly, it seems that for a fairly long time, there have been
two separate strands of development (Stewart [130]). One of them was the
study of positive definite functions, which started later but seems to have

20

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

been unaware of the fact that it considered a special case of positive definite
kernels. The latter was initiated by Hilbert [67] and Mercer [99], and was
pursued, for instance, by Schoenberg [115]. Hilbert calls a kernel k definit if
(50)

bZ b

k(x, x )f (x)f (x ) dx dx > 0

for all nonzero continuous functions fR, and shows that all eigenvalues of the
corresponding integral operator f 7 ab k(x, )f (x) dx are then positive. If k
R
satisfies the condition (50) subject to the constraint that ab f (x)g(x) dx = 0,
for some fixed function g, Hilbert calls it relativ definit. For that case, he
shows that k has at most one negative eigenvalue. Note that if f is chosen
to be constant, then this notion is closely related to the one of conditionally
positive definite kernels; see (17). For further historical details, see the review
of Stewart [130] or Berg at al. [18].
3. Convex programming methods for estimation. As we saw, kernels
can be used both for the purpose of describing nonlinear functions subject
to smoothness constraints and for the purpose of computing inner products
in some feature space efficiently. In this section we focus on the latter and
how it allows us to design methods of estimation based on the geometry of
the problems at hand.
Unless stated otherwise, E[] denotes the expectation with respect to all
random variables of the argument. Subscripts, such as EX [], indicate that
the expectation is taken over X. We will omit them wherever obvious. Finally, we will refer to Eemp [] as the empirical average with respect to an
n-sample. Given a sample S := {(x1 , y1 ), . . . , (xn , yn )} X Y, we now aim
at finding an affine function f (x) = hw, (x)i + b or in some cases a function f (x, y) = h(x, y), wi such that the empirical risk on S is minimized.
In the binary classification case this means that we want to maximize the
agreement between sgn f (x) and y.
Minimization of the empirical risk with respect to (w, b) is NP-hard (Minsky and Papert [101]). In fact, Ben-David et al. [15] show that even approximately minimizing the empirical risk is NP-hard, not only for linear
function classes but also for spheres and other simple geometrical objects.
This means that even if the statistical challenges could be solved, we still
would be confronted with a formidable algorithmic problem.
The indicator function {yf (x) < 0} is discontinuous and even small changes
in f may lead to large changes in both empirical and expected risk. Properties of such functions can be captured by the VC-dimension (Vapnik
and Chervonenkis [142]), that is, the maximum number of observations
which can be labeled in an arbitrary fashion by functions of the class.
Necessary and sufficient conditions for estimation can be stated in these

KERNEL METHODS IN MACHINE LEARNING

21

terms (Vapnik and Chervonenkis [143]). However, much tighter bounds

can be obtained by also using the scale of the class (Alon et al. [3]). In
fact, there exist function classes parameterized by a single scalar which
have infinite VC-dimension (Vapnik [140]).
Given the difficulty arising from minimizing the empirical risk, we now discuss algorithms which minimize an upper bound on the empirical risk, while
providing good computational properties and consistency of the estimators.
A discussion of the statistical properties follows in Section 3.6.
3.1. Support vector classification. Assume that S is linearly separable,
that is, there exists a linear function f (x) such that sgn yf (x) = 1 on S. In
this case, the task of finding a large margin separating hyperplane can be
viewed as one of solving (Vapnik and Lerner [145])
(51)

minimize 12 kwk2
w,b

subject to

yi (hw, xi + b) 1.

Note that kwk1 f (xi ) is the distance of the point xi to the hyperplane
H(w, b) := {x|hw, xi + b = 0}. The condition yi f (xi ) 1 implies that the
margin of separation is at least 2kwk1 . The bound becomes exact if equality
is attained for some yi = 1 and yj = 1. Consequently, minimizing kwk
subject to the constraints maximizes the margin of separation. Equation (51)
is a quadratic program which can be solved efficiently (Fletcher [51]).
Mangasarian [95] devised a similar optimization scheme using kwk1 instead of kwk2 in the objective function of (51). The result is a linear program. In general, one can show (Smola et al. [124]) that minimizing the p
norm of w leads to the maximizing of the margin of separation in the q
norm where p1 + 1q = 1. The 1 norm leads to sparse approximation schemes
(see also Chen et al. [29]), whereas the 2 norm can be extended to Hilbert
spaces and kernels.
To deal with nonseparable problems, that is, cases when (51) is infeasible,
we need to relax the constraints of the optimization problem. Bennett and
Mangasarian [17] and Cortes and Vapnik [34] impose a linear penalty on the
violation of the large-margin constraints to obtain
minimize 12 kwk2 + C
w,b,

(52)

n
X

i=1

subject to yi (hw, xi i + b) 1 i and i 0, i [n].

Equation (52) is a quadratic program which is always feasible (e.g., w, b = 0

and i = 1 satisfy the constraints). C > 0 is a regularization constant trading
off the violation of the constraints vs. maximizing the overall margin.
Whenever the dimensionality of X exceeds n, direct optimization of (52)
is computationally inefficient. This is particularly true if we map from X


T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

22

into an RKHS. To address these problems, one may solve the problem in
dual space as follows. The Lagrange function of (52) is given by
L(w, b, , , ) =

2
1
2 kwk

+C

n
X

i=1

(53)
+

n
X

i (1 i yi (hw, xi i + b))

i=1

n
X

i i ,

i=1

where i , i 0 for all i [n]. To compute the dual of L, we need to identify

the first order conditions in w, b. They are given by
w L = w

n
X

i yi xi = 0 and

i=1

b L =

(54)

n
X

i yi = 0

and

i=1

i L = C i + i = 0.
P

This translates into w = ni=1 i yi xi , the linear constraint ni=1 i yi = 0,

and the box-constraint i [0, C] arising from i 0. Substituting (54) into
L yields the Wolfe dual
minimize 12 Q 1

subject to y = 0 and i [0, C], i [n].

(55)
Q Rnn is the matrix of inner products Qij := yi yj hxi , xj i. Clearly, this can
be extended to feature maps and kernels easily via Kij := yi yj h(xi ), (xj )i =
yi yj k(xi , xj ). Note that w lies in the span of the xi . This is an instance of
the representer theorem (Theorem 9). The KKT conditions (Boser et al.
[23], Cortes and Vapnik [34], Karush [81] and Kuhn and Tucker [88]) require
that at optimality i (yi f (xi ) 1) = 0. This means that only those xi may
appear in the expansion (54) for which yi f (xi ) 1, as otherwise i = 0. The
xi with i > P
0 are commonly referred to as support vectors.
Note that ni=1 i is an upper bound on the empirical risk, as yi f (xi ) 0
implies i 1 (see also Lemma 10). The number of misclassified points xi
itself depends on the configuration of the data and the value of C. Ben-David
et al. [15] show that finding even an approximate minimum classification
error solution is difficult. That said, it is possible to modify (52) such that
a desired target number of observations violates yi f (xi ) for some
R by making the threshold itself a variable of the optimization problem
(Sch
olkopf et al. [120]). This leads to the following optimization problem
(-SV classification):
minimize 12 kwk2 +
w,b,

n
X
i=1

i n

KERNEL METHODS IN MACHINE LEARNING

(56)
subject to

23

yi (hw, xi i + b) i and i 0.

The dual of (56) is essentially identical to (55) with the exception of an

additional constraint:
minimize 12 Q

subject to

y = 0 and 1 = n and i [0, 1].

(57)
One can show that for every C there exists a such that the solution of
olkopf et al. [120] prove that
(57) is a multiple of the solution of (55). Sch
solving (57) for which > 0 satisfies the following:
1. is an upper bound on the fraction of margin errors.
2. is a lower bound on the fraction of SVs.
Moreover, under mild conditions, with probability 1, asymptotically, equals
both the fraction of SVs and the fraction of errors.
This statement implies that whenever the data are sufficiently well separable (i.e., > 0), -SV classification finds a solution with a fraction of at
most margin errors. Also note that, for = 1, all i = 1, that is, f becomes
an affine copy of the Parzen windows classifier (5).
3.2. Estimating the support of a density. We now extend the notion of
linear separation to that of estimating the support of a density (Sch
olkopf
et al. [117] and Tax and Duin [134]). Denote by X = {x1 , . . . , xn } X the
sample drawn from P(x). Let C be a class of measurable subsets of X and
let be a real-valued function defined on C. The quantile function (Einmal
and Mason [47]) with respect to (P, , C) is defined as
(58)

U () = inf{(C)|P(C) , C C}

where (0, 1].

We denote by C () and Cm () the (not necessarily unique) C C that

attain the infimum (when it is achievable) on P(x) and on the empirical
measure given by X respectively. A common choice of is the Lebesgue
measure, in which case C () is the minimum volume set C C that contains
at least a fraction of the probability mass.
Support estimation requires us to find some Cm () such that |P(Cm ())
| is small. This is where the complexity trade-off enters: On the one hand,
we want to use a rich class C to capture all possible distributions, on the
other hand, large classes lead to large deviations between and P(Cm ()).
Therefore, we have to consider classes of sets which are suitably restricted.
This can be achieved using an SVM regularizer.
SV support estimation works by using SV support estimation related
to previous work as follows: set (Cw ) = kwk2 , where Cw = {x|fw (x) },
fw (x) = hw, xi, and (w, ) are respectively a weight vector and an offset.


T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

24

Stated as a convex optimization problem, we want to separate the data

from the origin with maximum margin via
minimize 12 kwk2 +
w,,

n
X

i n

i=1

(59)

subject to

hw, xi i i and i 0.

Here, (0, 1] plays the same role as in (56), controlling the number of
observations xi for which f (xi ) . Since nonzero slack variables i are
penalized in the objective function, if w and solve this problem, then the
decision function f (x) will attain or exceed for at least a fraction 1 of
the xi contained in X, while the regularization term kwk will still be small.
The dual of (59) yield:
(60)

minimize 12 K subject to

1 = n and i [0, 1].

To compare (60) to a Parzen windows estimator, assume that k is such that

it can be normalized as a density in input space, such as a Gaussian. Using
= 1 in (60), the constraints automatically imply i = 1. Thus, f reduces to
a Parzen windows estimate of the underlying density. For < 1, the equality
constraint (60) still ensures that f is a thresholded density, now depending
only on a subset of Xthose which are important for deciding whether
f (x) .
3.3. Regression estimation. SV regression was first proposed in Vapnik
[140] and Vapnik et al. [144] using the so-called -insensitive loss function. It
is a direct extension of the soft-margin idea to regression: instead of requiring
that yf (x) exceeds some margin value, we now require that the values y
f (x) are bounded by a margin on both sides. That is, we impose the soft
constraints
yi f (xi ) i i

(61)

and f (xi ) yi i i ,

where i , i 0. If |yi f (xi )| , no penalty occurs. The objective function

is given by the sum of the slack variables i , i penalized by some C > 0 and
a measure for the slope of the function f (x) = hw, xi + b, that is, 21 kwk2 .
Before computing the dual of this problem, let us consider a somewhat
more general situation where we use a range of different convex penalties
for the deviation
between yi and f (xi ). One may check that minimizing
1
2 + C Pm + subject to (61) is equivalent to solving
kwk
i=1 i
i
2
(62) minimize 12 kwk2 +
w,b,

n
X

(yi f (xi ))

where () = max(0, || ).

i=1

Choosing different loss functions leads to a rather rich class of estimators:

KERNEL METHODS IN MACHINE LEARNING

25

() = 12 2 yields penalized least squares (LS) regression (Hoerl and Kennard [68], Morozov [102], Tikhonov [136] and Wahba [148]). The corresponding optimization problem can be minimized by solving a linear system.
For () = ||, we obtain the penalized least absolute deviations (LAD)
estimator (Bloomfield and Steiger [20]). That is, we obtain a quadratic
program to estimate the conditional median.
A combination of LS and LAD loss yields a penalized version of Hubers
olkopf [126]). In this case
robust regression (Huber [71] and Smola and Sch
1 2
we have () = 2
for || and () = || 2 for || .
Note that also quantile regression can be modified to work with kernels
(Sch
olkopf et al. [120]) by using as loss function the pinball loss, that
is, () = (1 ) if < 0 and () = if > 0.
All the optimization problems arising from the above five cases are convex
quadratic programs. Their dual resembles that of (61), namely,
(63a)
(63b)

1
minimize
2 ( ) K( ) + ( + ) y ( )

subject to

( ) 1 = 0 and i , i [0, C].

Here Kij = hxi , xj i for linear models and Kij = k(xi , xj ) if we map x (x).
The -trick, as described in (56) (Sch
olkopf et al. [120]), can be extended
to regression, allowing one to choose the margin of approximation automatically. In this case (63a) drops the terms in . In its place, we add a linear
constraint ( ) 1 = n. Likewise, LAD is obtained from (63) by dropping the terms in without additional constraints. Robust regression leaves
(63) unchanged, however, in the definition of K we have an additional term
of 1 on the main diagonal. Further details can be found in Sch
olkopf and
Smola [118]. For quantile regression we drop and we obtain different constants C(1 ) and C for the constraints on and . We will discuss
uniform convergence properties of the empirical risk estimates with respect
to various () in Section 3.6.
3.4. Multicategory classification, ranking and ordinal regression. Many
estimation problems cannot be described by assuming that Y = {1}. In
this case it is advantageous to go beyond simple functions f (x) depending on x only. Instead, we can encode a larger degree of information by
estimating a function f (x, y) and subsequently obtaining a prediction via
y(x) := arg maxyY f (x, y). In other words, we study problems where y is
obtained as the solution of an optimization problem over f (x, y) and we
wish to find f such that y matches yi as well as possible for relevant inputs
x.


T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

26

Note that the loss may be more than just a simple 01 loss. In the following
we denote by (y, y ) the loss incurred by estimating y instead of y. Without
loss of generality, we require that (y, y) = 0 and that (y, y ) 0 for all
y, y Y. Key in our reasoning is the following:
Lemma 10. Let f : X Y R and assume that (y, y ) 0 with
(y, y) = 0. Moreover, let 0 such that f (x, y) f (x, y ) (y, y )
for all y Y. In this case (y, arg maxy Y f (x, y )).
The construction of the estimator was suggested by Taskar et al. [132] and
Tsochantaridis et al. [137], and a special instance of the above lemma is given
by Joachims [78]. While the bound appears quite innocuous, it allows us to
describe a much richer class of estimation problems as a convex program.
To deal with the added complexity, we assume that f is given by f (x, y) =
h(x, y), wi. Given the possibly nontrivial connection between x and y, the
use of (x, y) cannot be avoided. Corresponding kernel functions are given
by k(x, y, x , y ) = h(x, y), (x , y )i. We have the following optimization
problem (Tsochantaridis et al. [137]):
minimize 12 kwk2 + C
w,

(64)
subject to

n
X

i=1

hw, (xi , yi ) (xi , y)i (yi , y) i , i [n], y Y.

This is a convex optimization problem which can be solved efficiently if the

constraints can be evaluated without high computational cost. One typically employs column-generation methods (Bennett et al. [16], Fletcher [51],
Hettich and Kortanek [66] and Tsochantaridis et al. [137]) which identify
one violated constraint at a time to find an approximate minimum of the
optimization problem.
To describe the flexibility of the framework set out by (64) we give several
examples below:
Binary classification can be recovered by setting (x, y) = y(x), in which
case the constraint of (64) reduces to 2yi h(xi ), wi 1 i . Ignoring constant offsets and a scaling factor of 2, this is exactly the standard SVM
optimization problem.
Multicategory classification problems (Allwein et al. [2], Collins [30] and
Crammer and Singer [35]) can be encoded via Y = [N ], where N is the
number of classes and (y, y ) = 1 y,y . In other words, the loss is 1
whenever we predict the wrong class and 0 for correct classification. Corresponding kernels are typically chosen to be y,y k(x, x ).
We can deal with joint labeling problems by setting Y = {1}n . In other
words, the error measure does not depend on a single observation but on

KERNEL METHODS IN MACHINE LEARNING

27

an entire set of labels. Joachims [78] shows that the so-called F1 score
(van Rijsbergen [138]) used in document retrieval and the area under the
ROC curve (Bamber [10]) fall into this category of problems. Moreover,
Joachims [78] derives an O(n2 ) method for evaluating the inequality constraint over Y.
Multilabel estimation problems deal with the situation where we want to
find the best subset of labels Y 2[N ] which correspond to some observation x. Elisseeff and Weston [48] devise a ranking scheme where
f (x, i) > f (x, j) if label i y and j
/ y. It is a special case of an approach
described next.
Note that (64) is invariant under translations (x, y) (x, y) + 0 where
0 is constant, as (xi , yi ) (xi , y) remains unchanged. In practice, this
means that transformations k(x, y, x , y ) k(x, y, x , y ) + h0 , (x, y)i +
h0 , (x , y )i + k0 k2 do not affect the outcome of the estimation process.
Since 0 was arbitrary, we have the following lemma:
Lemma 11. Let H be an RKHS on X Y with kernel k. Moreover, let
g H. Then the function k(x, y, x , y ) + f (x, y) + f (x, y ) + kgk2H is a kernel
and it yields the same estimates as k.
We need a slight extension to deal with general ranking problems. Denote
by Y = Graph[N ] the set of all directed graphs on N vertices which do not
contain loops of less than three nodes. Here an edge (i, j) y indicates that
i is preferred to j with respect to the observation x. It is the goal to find
some function f : X [N ] R which imposes a total order on [N ] (for a
given x) by virtue of the function values f (x, i) such that the total order
and y are in good agreement.
More specifically, Crammer and Singer [36] and Dekel et al. [45] propose a
decomposition algorithm A for the graphs y such that the estimation error
is given by the number of subgraphs of y which are in disagreement with
the total order imposed by f . As an example, multiclass classification can
be viewed as a graph y where the correct label i is at the root of a directed
graph and all incorrect labels are its children. Multilabel classification is
then a bipartite graph where the correct labels only contain outgoing arcs
and the incorrect labels only incoming ones.
This setting leads to a form similar to (64), except for the fact that we
now have constraints over each subgraph G A(y). We solve
minimize 21 kwk2 + C
w,

(65)

subject to

n
X
i=1

|A(yi )|1

iG

GA(yi )

hw, (xi , u) (xi , v)i 1 iG and iG 0

for all (u, v) G A(yi ).

28

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

That is, we test for all (u, v) G whether the ranking imposed by G yi is
satisfied.
Finally, ordinal regression problems which perform ranking not over labels y but rather over observations x were studied by Herbrich et al. [65] and
Chapelle and Harchaoui [27] in the context of ordinal regression and conjoint
analysis respectively. In ordinal regression x is preferred to x if f (x) > f (x )
and, hence, one minimizes an optimization problem akin to (64), with constraint hw, (xi ) (xj )i 1 ij . In conjoint analysis the same operation
is carried out for (x, u), where u is the user under consideration. Similar
models were also studied by Basilico and Hofmann [13]. Further models will
be discussed in Section 4, in particular situations where Y is of exponential
size. These models allow one to deal with sequences and more sophisticated
structures.
3.5. Applications of SVM algorithms. When SVMs were first presented,
they initially met with skepticism in the statistical community. Part of the
reason was that, as described, SVMs construct their decision rules in potentially very high-dimensional feature spaces associated with kernels. Although
there was a fair amount of theoretical work addressing this issue (see Section 3.6 below), it was probably to a larger extent the empirical success of
SVMs that paved its way to become a standard method of the statistical
toolbox. The first successes of SVMs on practical problems were in handwritten digit recognition, which was the main benchmark task considered in the
Adaptive Systems Department at AT&T Bell Labs where SVMs were developed. Using methods to incorporate transformation invariances, SVMs were
shown to beat the world record on the MNIST benchmark set, at the time
the gold standard in the field (DeCoste and Sch
olkopf [44]). There has been
a significant number of further computer vision applications of SVMs since
then, including tasks such as object recognition and detection. Nevertheless,
it is probably fair to say that two other fields have been more influential in
spreading the use of SVMs: bioinformatics and natural language processing.
Both of them have generated a spectrum of challenging high-dimensional
problems on which SVMs excel, such as microarray processing tasks and
text categorization. For references, see Joachims [77] and Sch
olkopf et al.
[121].
Many successful applications have been implemented using SV classifiers;
however, also the other variants of SVMs have led to very good results,
including SV regression, SV novelty detection, SVMs for ranking and, more
recently, problems with interdependent labels (McCallum et al. [96] and
Tsochantaridis et al. [137]).
At present there exists a large number of readily available software packages for SVM optimization. For instance, SVMStruct, based on Tsochantaridis et al. [137] solves structured estimation problems. LibSVM is an

KERNEL METHODS IN MACHINE LEARNING

29

open source solver which excels on binary problems. The Torch package
contains a number of estimation methods, including SVM solvers. Several
SVM implementations are also available via statistical packages, such as R.
3.6. Margins and uniform convergence bounds. While the algorithms
were motivated by means of their practicality and the fact that 01 loss
functions yield hard-to-control estimators, there exists a large body of work
on statistical analysis. We refer to the works of Bartlett and Mendelson [12],
Jordan et al. [80], Koltchinskii [86], Mendelson [98] and Vapnik [141] for
details. In particular, the review of Bousquet et al. [24] provides an excellent summary of the current state of the art. Specifically for the structured
case, recent work by Collins [30] and Taskar et al. [132] deals with explicit
constructions to obtain better scaling behavior in terms of the number of
class labels.
The general strategy of the analysis can be described by the following
three steps: first, the discrete loss is upper bounded by some function, such
as (yf (x)), which can be efficiently minimized [e.g. the soft margin function
max(0, 1 yf (x)) of the previous section satisfies this property]. Second, one
proves that the empirical average of the -loss is concentrated close to its
expectation. This will be achieved by means of Rademacher averages. Third,
one shows that under rather general conditions the minimization of the loss is consistent with the minimization of the expected risk. Finally, these
bounds are combined to obtain rates of convergence which only depend on
the Rademacher average and the approximation properties of the function
class under consideration.
4. Statistical models and RKHS. As we have argued so far, the reproducing kernel Hilbert space approach offers many advantages in machine
learning: (i) powerful and flexible models can be defined, (ii) many results
and algorithms for linear models in Euclidean spaces can be generalized
to RKHS, (iii) learning theory assures that effective learning in RKHS is
possible, for instance, by means of regularization.
In this chapter we will show how kernel methods can be utilized in the
context of statistical models. There are several reasons to pursue such an avenue. First of all, in conditional modeling, it is often insufficient to compute a
prediction without assessing confidence and reliability. Second, when dealing
with multiple or structured responses, it is important to model dependencies between responses in addition to the dependence on a set of covariates.
Third, incomplete data, be it due to missing variables, incomplete training
targets or a model structure involving latent variables, needs to be dealt
with in a principled manner. All of these issues can be addressed by using
the RKHS approach to define statistical models and by combining kernels
with statistical approaches such as exponential models, generalized linear
models and Markov networks.


T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

30

4.1. Exponential RKHS models.

4.1.1. Exponential models. Exponential models or exponential families
are among the most important class of parametric models studied in statistics. Given a canonical vector of statistics and a -finite measure over
the sample space X , an exponential model can be defined via its probability
density with respect to (cf. Barndorff-Nielsen [11]),
p(x; ) = exp[h, (x)i g()]
(66)
where g() := ln

eh,(x)i d(x).

X
m
R : g() < }

The m-dimensional vector with := {

is also called
the canonical parameter vector. In general, there are multiple exponential
representations of the same model via canonical parameters that are affinely
related to one another (Murray and Rice [103]). A representation with minimal m is called a minimal representation, in which case m is the order of
the exponential model. One of the most important properties of exponential
families is that they have sufficient statistics of fixed dimensionality, that is,
the joint density for i.i.d. random variables X1 , X2 , . . . , XnPis also exponential, the corresponding canonical statistics simply being ni=1 (Xi ). It is
well known that much of the structure of exponential models can be derived
from the log partition function g(), in particular,
(67)

g() = () := E [(X)],

2 g() = V [(X)],

where is known as the mean-value map. Being a covariance matrix, the

Hessian of g is positive semi-definite and, consequently, g is convex.
Maximum likelihood estimation (MLE) in exponential families leads to
a particularly elegant form for the MLE equations: the expected and the
This means, given an
observed canonical statistics agree at the MLE .
i.i.d. sample S = (xi )i[n] ,
(68)

n
1X

E[(X)] = () =
(xi ) := rES [(X)].
n i=1

4.1.2. Exponential RKHS models. One can extend the parameteric exponential model in (66) by defining a statistical model via an RKHS H
with generating kernel k. Linear function h, ()i over X are replaced with
functions f H, which yields an exponential RKHS model
(69)

p(x; f ) = exp[f (x) g(f )],

(

f H := f : f () =

xS

x k(, x), S X , |S| < .

31

KERNEL METHODS IN MACHINE LEARNING

A justification for using exponential RKHS families with rich canonical

statistics as a generic way to define nonparametric models stems from the
fact that if the chosen kernel k is powerful enough, the associated exponential
families become universal density estimators. This can be made precise using
the concept of universal kernels (Steinwart [128], cf. Section 2).
Proposition 12 (Dense densities). Let X be a measurable set with a
fixed -finite measure and denote by P a family of densities on X with
respect to such that p P is uniformly bounded from above and continuous.
Let k : X X R be a universal kernel for H. Then the exponential RKHS
family of densities generated by k according to equation (69) is dense in P
in the L sense.
4.1.3. Conditional exponential models. For the rest of the paper we will
focus on the case of predictive or conditional modeling with apotentially
compound or structuredresponse variable Y and predictor variables X.
Taking up the concept of joint kernels introduced in the previous section, we
will investigate conditional models that are defined by functions f : X Y
R from some RKHS H over X Y with kernel k as follows:
p(y|x; f ) = exp[f (x, y) g(x, f )]
(70)
where g(x, f ) := ln

ef (x,y) d(y).

Notice that in the finite-dimensional case we have a feature map : X Y

Rm from which parametric models are obtained via H := {f : w, f (x, y) =
f (x, y; w) := hw, (x, y)i} and each f can be identified with its parameter
w. Let us discuss some concrete examples to illustrate the rather general
model equation (70):
Let Y be univariate and define (x, y) = y(x). Then simply f (x, y; w) =
hw, (x, y)i = y f(x; w), with f(x; w) := hw, (x)i and the model equation
in (70) reduces to
(71)

p(y|x; w) = exp[yhw, (x)i g(x, w)].

This is a generalized linear model (GLM) (McCullagh and Nelder [97])

with a canonical link, that is, the canonical parameters depend linearly
on the covariates (x). For different response scales, we get several wellknown models such as, for instance, logistic regression where y {1, 1}.
In the nonparameteric extension of generalized linear models following
Green and Yandell [57] and OSullivan [105] the parametric assumption
on the linear predictor f(x; w) = hw, (x)i in the GLMs is relaxed by
requiring that f comes from some sufficiently smooth class of functions,

32

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

namely, an RKHS defined over X . In combination with a parametric part,

this can also be used to define semi-parametric models. Popular choices of
kernels include the ANOVA kernel investigated by [149]. This is a special
case of defining joint kernels from an existing kernel k over inputs via
k((x, y), (x , y )) := yy k(x, x ).
Joint kernels provide a powerful framework for prediction problems with
structured outputs. An illuminating example is statistical natural language parsing with lexicalized probabilistic context free grammars (Magerman [94]). Here x will be an English sentence and y a parse tree for x,
that is, a highly structured and complex output. The productions of the
grammar are known, but the conditional probability p(y|x) needs to be
estimated based on training data of parsed/annotated sentences. In the
simplest case, the extracted statistics may encode the frequencies of the
use of different productions in a sentence with a known parse tree. More
sophisticated feature encodings are discussed in Taskar et al. [133] and
Zettlemoyer and Collins [156]. The conditional modeling approach provide alternatives to state-of-the art approaches that estimate joint models
p(x, y) with maximum likelihood or maximum entropy and obtain predictive models by conditioning on x.
4.1.4. Risk functions for model fitting. There are different inference principles to determine the optimal function f H for the conditional exponential model in (70). One standard approach to parametric model fitting is to
maximize the conditional log-likelihoodor equivalentlyminimize a logarithmic loss, a strategy pursued in the Conditional Random Field (CRF)
approach of Lafferty [90]. Here we consider the more general case of minimizing a functional that includes a monotone function of the Hilbert space norm
kf kH as a stabilizer (Wahba [148]). This reduces to penalized log-likelihood
estimation in the finite-dimensional case,
C ll (f ; S) :=
(72)

n
1X
ln p(yi |xi ; f ),
n i=1

fll (S) := arg min kf k2H + C ll(f ; S).

f H 2

For the parametric case, Lafferty et al. [90] have employed variants of
improved iterative scaling (Darroch and Ratcliff [40] and Della Pietra
[46]) to optimize equation (72), whereas Sha and Pereira [122] have investigated preconditioned conjugate gradient descent and limited memory
quasi-Newton methods.
In order to optimize equation (72) one usually needs to compute expectations of the canonical statistics Ef [(Y, x)] at sample points x = xi , which
requires the availability of efficient inference algorithms.

33

KERNEL METHODS IN MACHINE LEARNING

As we have seen in the case of classification and regression, likelihoodbased criteria are by no means the only justifiable choice and large margin
methods offer an interesting alternative. To that extend, we will present
a general formulation of large margin methods for response variables over
finite sample spaces that is based on the approach suggested by Altun et al.
[6] and Taskar et al. [132]. Define
r(x, y; f ) := f (x, y) max
f (x, y ) = min
log

y 6=y

(73)

y 6=y

p(y|x; f )
p(y |x; f )

and

r(S; f ) := min r(xi , yi ; f ).

i=1

Here r(S; f ) generalizes the notion of separation margin used in SVMs.

Since the log-odds ratio is sensitive to rescaling of f , that is, r(x, y; f ) =
r(x, y; f ), we need to constrain kf kH to make the problem well defined. We
thus replace f by 1 f for some fixed dispersion parameter > 0 and define
the maximum margin problem fmm (S) := 1 arg maxkf kH =1 r(S; f /). For
the sake of the presentation, we will drop in the following. (We will not
deal with the problem of how to estimate here; note, however, that one
does need to know in order to make an optimal deterministic prediction.)
Using the same line of arguments as was used in Section 3, the maximum
margin problem can be re-formulated as a constrained optimization problem
(74)

fmm (S) := arg min 12 kf k2H

f H

s.t. r(xi , yi ; f ) 1, i [n],

provided the latter is feasible, that is, if there exists f H such that r(S; f ) >
0. To make the connection to SVMs, consider the case of binary classification with (x, y) = y(x), f (x, y; w) = hw, y(x)i, where r(x, y; f ) =
hw, y(x)i hw, y(x)i = 2yhw, (x)i = 2(x, y; w). The latter is twice the
standard margin for binary classification in SVMs.
A soft margin version can be defined based on the Hinge loss as follows:
n
1X
C (f ; S) :=
min{1 r(xi , yi ; f ), 0},
n i=1
hl

(75)

fsm (S) := arg min kf k2H + C hl (f, S).

f H 2

An equivalent formulation using slack variables i as discussed in Section 3

can be obtained by introducing soft-margin constraints r(xi , yi ; f ) 1 i ,
i 0 and by defining C hl = n1 i . Each nonlinear constraint can be further
expanded into |Y| linear constraints f (xi , yi ) f (xi , y) 1 i for all
y 6= yi .


T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

34

Prediction problems with structured outputs often involve task-specific

loss function : Y Y R discussed in Section 3.4. As suggested in
Taskar et al. [132] cost sensitive large margin methods can be obtained by
defining re-scaled margin constraints f (xi , yi ) f (xi , y) (yi , y) i .
Another sensible option in the parametric case is to minimize an exponential risk function of the following type:
(76)

n X
X

1
exp[f (xi , yi ; w) f (xi , y; w)].
fexp (S) := arg min
n i=1 y6=y
w
i

This is related to the exponential loss used in the AdaBoost method of

Freund and Schapire [53]. Since we are mainly interested in kernel-based
methods here, we refrain from further elaborating on this connection.
4.1.5. Generalized representer theorem and dual soft-margin formulation.
It is crucial to understand how the representer theorem applies in the setting of arbitrary discrete output spaces, since a finite representation for the
optimal f {fll , fsm } is the basis for constructive model fitting. Notice that
the regularized log-loss, as well as the soft margin functional introduced
above, depends not only on the values of f on the sample S, but rather on
the evaluation of f on the augmented sample S := {(xi , y) : i [n], y Y}.
This is the case, because for each xi , output values y 6= yi not observed with
xi show up in the log-partition function g(xi , f ) in (70), as well as in the
log-odds ratios in (73). This adds an additional complication compared to
binary classification.
Corollary 13. Denote by H an RKHS on X Y with kernel k and
let S = ((xi , yi ))i[n] . Furthermore, let C(f ; S) be a functional depending
Let be a strictly
on f only via its values on the augmented sample S.
monotonically increasing function. Then the solution of the optimization
+ (kf kH ) can be written as
problem f(S) := arg minf H C(f ; S)
(77)

f() =

n X
X

iy k(, (xi , y)).

i=1 yY

This follows directly from Theorem 9.

Let us focus on the soft margin maximizer fsm . Instead of solving (75)
directly, we first derive the dual program, following essentially the derivation
in Section 3.
Proposition 14 (Tsochantaridis et al. [137]). The minimizer fsm (S)
can be written as in Corollary 13, where the expansion coefficients can be

35

KERNEL METHODS IN MACHINE LEARNING

computed from the solution of the following convex quadratic program:

(78a)
(78b)

= arg min

s.t.

1
2

n X X
X

iy jy Kiy,jy

i,j=1 y6=yi y 6=yj

n X
X

iy

i=1 y6=yi

iy 1, i [n]; iy 0, i [n], y Y,

y6=yi

where Kiy,jy := k((xi , yi ), (xj , yj )) + k((xi , y), (xj , y )) k((xi , yi ), (xj , y ))

k((xi , y), (xj , yj )).
The multiclass SVM formulation of [35] can be recovered as a special
case for kernels that are diagonal with respect to the outputs, that is,
k((x, y), (x , y )) = y,y k(x, x ). Notice that in this case the quadratic part
in equation (78a) simplifies to
X
i,j

k(xi , xj )

iy jy [1 + yi ,y yj ,y yi ,y yj ,y ].

The pairs (xi , y) for which iy > 0 are the support pairs, generalizing
the notion of support vectors. As in binary SVMs, their number can be
much smaller than the total number of constraints. Notice also that in the
final expansion contributions k(, (xi , yi )) will get nonnegative weights,
whereas k(, (xi , y)) for y 6= yi P
will get nonpositive weights. Overall one
gets a balance equation iyi y6=yi iy = 0 for every data point.

4.1.6. Sparse approximation. Proposition 14 shows that sparseness in

the representation of fsm is linked to the fact that only few iy in the solution to the dual problem in equation (78) are nonzero. Note that each of
these Lagrange multipliers is linked to the corresponding soft margin constraints f (xi , yi ) f (xi , y) 1 i . Hence, sparseness is achieved, if only few
constraints are active at the optimal solution. While this may or may not be
the case for a given sample, one can still exploit this observation to define
a nested sequence of relaxations, where margin constraint are incrementally
added. This corresponds to a constraint selection algorithm (Bertsimas and
Tsitsiklis [19]) for the primal or, equivalently, a variable selection or column generation method for the dual program and has been investigated in
Tsochantaridis et al. [137]. Solving a sequence of increasingly tighter relaxations to a mathematical problem is also known as an outer approximation.
In particular, one may iterate through the training examples according to
some (fair) visitation schedule and greedily select constraints that are most
violated at the current solution f , that is, for the ith instance one computes
(79)

yi = arg max f (xi , y) = arg max p(y|xi ; f ),

y6=yi

y6=yi

36

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

and then strengthens the current relaxation by including iyi in the optimization of the dual if f (xi , yi ) f (xi , yi ) < 1 i . Here > 0 is a
pre-defined tolerance parameter. It is important to understand how many
strengthening steps are necessary to achieve a reasonable close approximation to the original problem. The following theorem provides an answer:
= maxi,y Kiy,iy and
Theorem 15 (Tsochantaridis et al. [137]). Let R
choose > 0. A sequential strengthening procedure, which optimizes equation (75) by greedily selecting -violated constraints, will find an approximate solution where all constraints are fulfilled within a precision of , that
2
4R
is, r(xi , yi ; f ) 1 i after at most 2n
max{1, n2 } steps.
the optimal solution of a relaxation
Corollary 16. Denote by (f, )
of the problem in Proposition 14, minimizing R(f, , S) while violating no
constraint by more than (cf. Theorem 15). Then
S) R(fsm , , S) R(f, ,
S) + ,
R(f, ,
where (fsm , ) is the optimal solution of the original problem.
Combined with an efficient QP solver, the above theorem guarantees a
and 1 . This holds irrespective of speruntime polynomial in n, 1 , R
cial properties of the data set utilized, the only exception being the de
pendency on the sample points xi is through the radius R.
The remaining key problem is how to compute equation (79) efficiently.
The answer depends on the specific form of the joint kernel k and/or
the feature map . In many cases, efficient dynamic programming techniques exists, whereas in other cases one has to resort to approximations
or use other methods to identify a set of candidate distractors Yi Y for
a training pair (xi , yi ) (Collins [30]). Sometimes one may also have search
heuristics available that may not find the solution to Equation (79), but
that find (other) -violating constraints with a reasonable computational
effort.
4.1.7. Generalized Gaussian processes classification. The model
equation (70) and the minimization of the regularized log-loss can be interpreted as a generalization of Gaussian process classification (Altun et al.
[4] and Rasmussen and Williams [111]) by assuming that (f (x, ))xX is a
vector-valued zero mean Gaussian process; note that the covariance function
C is defined over pairs X Y. For a given sample S, define a multi-index
vector F (S) := (f (xi , y))i,y as the restriction of the stochastic process f to
Denote the kernel matrix by K = (Kiy,jy ), where
the augmented sample S.
Kiy,jy := C((xi , y), (xj , y )) with indices i, j [n] and y, y Y, so that, in

37

KERNEL METHODS IN MACHINE LEARNING

summary, F (S) N (0, K). This induces a predictive model via Bayesian
model integration according to
p(y|x; S) =

(80)

p(y|F (x, ))p(F |S) dF,

where x is a test point that has been included in the sample (transductive
setting). For an i.i.d. sample, the log-posterior for F can be written as
(81)

ln p(F |S) = 12 F T K1 F +

n
X

[f (xi , yi ) g(xi , F )] + const.

i=1

Invoking the representer theorem for F (S) := arg maxF ln p(F |S), we know
that
F (S)iy =

(82)

n X
X

iy Kiy,jy ,

j=1 y Y

which we plug into equation (81) to arrive at

(83)

min T K

n
X
i=1

T Keiyi + log

yY

exp[T Keiy ] ,
P

where eiy denotes the respective unit vector. Notice that for f () = i,y iy k(,
(xi , y)) thePfirstPterm is equivalent to the squared RKHS norm of f H since
hf, f iH = i,j y,y iy jy hk(, (xi , y)), k(, (xj , y ))i. The latter inner product reduces to k((xi , y), (xj , y )) due to the reproducing property. Again, the
key issue in solving (83) is how to achieve spareness in the expansion for F .

4.2. Markov networks and kernels. In Section 4.1 no assumptions about

the specific structure of the joint kernel defining the model in equation (70)
has been made. In the following, we will focus on a more specific setting
with multiple outputs, where dependencies are modeled by a conditional
independence graph. This approach is motivated by the fact that independently predicting individual responses based on marginal response models
will often be suboptimal and explicitly modeling these interactions can be
of crucial importance.
4.2.1. Markov networks and factorization theorem. Denote predictor variables by X, response variables by Y and define Z := (X, Y ) with associated
sample space Z. We use Markov networks as the modeling formalism for representing dependencies between covariates and response variables, as well as
interdependencies among response variables.

38

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

Definition 17. A conditional independence graph (or Markov network)

is an undirected graph G = (Z, E) such that for any pair of variables (Zi , Zj )
/
E if and only if Zi
Zj |Z {Zi , Zj }.
The above definition is based on the pairwise Markov property, but by
virtue of the separation theorem (see, e.g., Whittaker [154]) this implies
the global Markov property for distributions with full support. The global
Markov property says that for disjoint subsets U, V, W Z where W separates U from V in G one has that U
V |W . Even more important in
the context of this paper is the factorization result due to Hammersley and
Clifford [61].
Theorem 18. Given a random vector Z with conditional independence
graph G, any density function for Z with full support factorizes over C (G),
the set of maximal cliques of G as follows:
(84)

p(z) = exp

"

cC (G)

fc (zc ) ,

where fc are clique compatibility functions dependent on z only via the restriction on clique configurations zc .
The significance of this result is that in order to specify a distribution for
Z, one only needs to specify or estimate the simpler functions fc .
4.2.2. Kernel decomposition over Markov networks. It is of interest to
analyze the structure of kernels k that generate Hilbert spaces H of functions
that are consistent with a graph.
Definition 19. A function f : Z R is compatible withPa conditional
independence graph G, if f decomposes additively as f (z) = cC (G) fc (zc )
with suitably chosen functions fc . A Hilbert space H over Z is compatible
with G, if every function f H is compatible with G. Such f and H are also
called G-compatible.
Proposition 20. Let H with kernel k be a G-compatible RKHS. Then
there are functions kcd : Zc Zd R such that the kernel decomposes as
k(u, z) =

kcd (uc , zd ).

c,dC

Lemma 21. Let X be a set of n-tupels and fi , gi : X X P

R for i [n]
functions
such
that
f
(x,
y)
=
f
(x
,
y)
and
g
(x,
y)
=
g
(x,
y
).
If
f (x , y) =
i
i
i
i
i
i
Pi i i
P
f
g
(x,
y
)
for
all
x,
y,
then
there
exist
functions
h
such
that
j
ij
i i (xi , y) =
Pj j
h
(x
,
y
).
i,j ij i j

KERNEL METHODS IN MACHINE LEARNING

39

Proposition 20 is useful for the design of kernels, since it states that only
kernels allowing an additive decomposition into local functions kcd are
compatible with a given Markov network G. Lafferty et al. [89] have pursued a similar approach by considering kernels for RKHS with functions
defined over ZC := {(c, zc ) : c c, zc Zc }. In the latter case one can even
deal with cases where the conditional dependency graph is (potentially)
different for every instance.
An illuminating example of how to design kernels via the decomposition in Proposition 20 is the case of conditional Markov chains, for which
models based on joint kernels have been proposed in Altun et al. [6],
Collins [30], Lafferty et al. [90] and Taskar et al. [132]. Given an input
sequences X = (Xt )t[T ] , the goal is to predict a sequence of labels or
class variables Y = (Yt )t[T ] , Yt . Dependencies between class variables
are modeled in terms of a Markov chain, whereas outputs Yt are assumed
to depend (directly) on an observation window (Xtr , . . . , Xt , . . . , Xt+r ).
Notice that this goes beyond the standard hidden Markov model structure by allowing for overlapping features (r 1). For simplicity, we focus on a window size of r = 1, in which case the clique set is given by
C := {ct := (xt , yt , yt+1 ), ct := (xt+1 , yt , yt+1 ) : t [T 1]}. We assume an
input kernel k is given and introduce indicator vectors (or dummy variates) I(Y{t,t+1} ) := (I, (Y{t,t+1} )), . Now we can define the local kernel functions as

kcd (zc , zd ) := hI(y{s,s+1} )I(y{t,t+1}

)i

(85)

k(xs , xt ),
k(xs+1 , xt+1 ),

if c = cs and d = ct ,
if c = cs and d = ct .

Notice that the inner product between indicator vectors is zero, unless the
variable pairs are in the same configuration.
Conditional Markov chain models have found widespread applications
in natural language processing (e.g., for part of speech tagging and shallow
parsing, cf. Sha and Pereira [122]), in information retrieval (e.g., for information extraction, cf. McCallum et al. [96]) or in computational biology
(e.g., for gene prediction, cf. Culotta et al. [39]).
4.2.3. Clique-based sparse approximation. Proposition 20 immediately
leads to an alternative version of the representer theorem as observed by
Lafferty et al. [89] and Altum et al. [4].
Corollary 22. If H is G-compatible then in the same setting as in
Corollary 13, the optimizer f can be written as
(86)

f(u) =

n X X
X

i=1 cC yc Yc

i
c,y
c

dC

kcd ((xic , yc ), ud ),

40

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

here xic are the variables of xi belonging to clique c and Yc is the subspace
of Zc that contains response variables.
Notice that thePnumber of parameters in the representation equation (86)
scales with n cC |Yc | as opposed to n |Y| in equation (77). For cliques
with reasonably small state spaces, this will be a significantly more compact representation. Notice also that the evaluation of functions kcd will
typically be more efficient than evaluating k.
In spite of this improvement, the number of terms in the expansion in
equation (86) may in practice still be too large. In this case, one can pursue
a reduced set approach, which selects a subset of variables to be included
in a sparsified expansion. This has been proposed in Taskar et al. [132] for
the soft margin maximization problem, as well as in Altun et al. [5] and
Lafferty et al. [89] for conditional random fields and Gaussian processes.
i
For instance, in Lafferty et al. [89] parameters cy
that maximize the
c
functional gradient of the regularized log-loss are greedily included in the
reduced set. In Taskar et al. [132] a similar selection criterion is utilized
with respect to margin violations, leading to an SMO-like optimization
algorithm (Platt [107]).
4.2.4. Probabilistic inference. In dealing with structured or interdependent response variables, computing marginal probabilities of interest or computing the most probable response [cf. equation (79)] may be nontrivial.
However, for dependency graphs with small tree width, efficient inference
algorithms exist, such as the junction tree algorithm (Dawid [43] and Jensen
et al. [76]) and variants thereof. Notice that in the case of the conditional
or hidden Markov chain, the junction tree algorithm is equivalent to the
well-known forwardbackward algorithm (Baum [14]). Recently, a number
of approximate inference algorithms have been developed to deal with dependency graphs for which exact inference is not tractable (see, e.g., Wainwright
and Jordan [150]).
5. Kernel methods for unsupervised learning. This section discusses various methods of data analysis by modeling the distribution of data in feature
space. To that extent, we study the behavior of (x) by means of rather simple linear methods, which have implications for nonlinear methods on the
original data space X . In particular, we will discuss the extension of PCA to
Hilbert spaces, which allows for image denoising, clustering, and nonlinear
dimensionality reduction, the study of covariance operators for the measure
of independence, the study of mean operators for the design of two-sample
tests, and the modeling of complex dependencies between sets of random
variables via kernel dependency estimation and canonical correlation analysis.

KERNEL METHODS IN MACHINE LEARNING

41

5.1. Kernel principal component analysis. Principal component analysis (PCA) is a powerful technique for extracting structure from possibly
high-dimensional data sets. It is readily performed by solving an eigenvalue
problem, or by using iterative algorithms which estimate principal components.
PCA is an orthogonal transformation of the coordinate system in which
we describe our data. The new coordinate system is obtained by projection
onto the so-called principal axes of the data. A small number of principal
components is often sufficient to account for most of the structure in the
data.
The basic idea is strikingly simple: denote by X = {x1 , . . . , xn } an nsample drawn from P(x). Then the covariance operator C is given by C =
E[(x E[x])(x E[x]) ]. PCA aims at estimating leading eigenvectors of
C via the empirical estimate Cemp = Eemp [(x Eemp [x])(x Eemp [x]) ]. If
X is d-dimensional, then the eigenvectors can be computed in O(d3 ) time
(Press et al. [110]).
The problem can also be posed in feature space (Sch
olkopf et al. [119])
by replacing x with (x). In this case, however, it is impossible to compute the eigenvectors directly. Yet, note that the image of Cemp lies in the
span of {(x1 ), . . . , (xn )}. Hence, it is sufficient to diagonalize Cemp in
that subspace. In other words, we replace the outer product Cemp by an inner product matrix, leaving P
the eigenvalues unchanged, which can be computed efficiently. Using w = ni=1 i (xi ), it follows that needs to satisfy
P KP = , where P is the projection operator with Pij = ij n2 and
K is the kernel matrix on X.
Note that the problem can also be recovered as one of maximizing some
Contrast[f, X] subject to f F . This means that the projections onto the
leading eigenvectors correspond to the most reliable features. This optimization problem also allows us to unify various feature extraction methods as
follows:
For Contrast[f, X] = Varemp [f, X] and F = {hw, xi subject to kwk 1},
we recover PCA.
Changing F to F = {hw, (x)i subject to kwk 1}, we recover kernel
PCA.
For Contrast[f, X] = Curtosis[f, X] and F = {hw, xi subject to kwk 1},
we have Projection Pursuit (Friedman and Tukey [55] and Huber [72]).
Other contrasts lead to further variants, that is, the Epanechikov kernel,
entropic contrasts, and so on (Cook et al. [32], Friedman [54] and Jones
and Sibson [79]).
If F is a convex combination of basis functions and the contrast function
is convex in w, one obtains computationally efficient algorithms, as the
solution of the optimization problem can be found at one of the vertices
olkopf and Smola [118]).
(Rockafellar [114] and Sch

42

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

Subsequent projections are obtained, for example, by seeking directions orthogonal to f or other computationally attractive variants thereof.
Kernel PCA has been applied to numerous problems, from preprocessing and invariant feature extraction (Mika et al. [100]) to image denoising
and super-resolution (Kim et al. [84]). The basic idea in the latter case is
to obtain a set of principal directions in feature space w1 , . . . , wl , obtained
from noise-free data, and to project the image (x) of a noisy observation x

onto the space spanned by w1 , . . . , wl . This yields a denoised solution (x)

in feature space. Finally, to obtain the pre-image of this denoised solution,

one minimizes k(x ) (x)k.

The fact that projections onto the leading
principal components turn out to be good starting points for pre-image iterations is further exploited in kernel dependency estimation (Section 5.3).
Kernel PCA can be shown to contain several popular dimensionality reduction algorithms as special cases, including LLE, Laplacian Eigenmaps and
(approximately) Isomap (Ham et al. [60]).
5.2. Canonical correlation and measures of independence. Given two samples X, Y , canonical correlation analysis (Hotelling [70]) aims at finding directions of projection u, v such that the correlation coefficient between X
and Y is maximized. That is, (u, v) are given by
arg max Varemp [hu, xi]1 Varemp [hv, yi]1
(87)

u,v

Eemp [hu, x Eemp [x]ihv, y Eemp [y]i].

1/2

1/2

This problem can be solved by finding the eigensystem of Cx Cxy Cy

,
where Cx , Cy are the covariance matrices of X and Y and Cxy is the covariance matrix between X and Y , respectively. Multivariate extensions are
discussed in Kettenring [83].
CCA can be extended to kernels by means of replacing linear projections
hu, xi by projections in feature space hu, (x)i. More specifically, Bach and
Jordan [8] used the so-derived contrast to obtain a measure of independence
and applied it to Independent Component Analysis with great success. However, the formulation requires an additional regularization term to prevent
the resulting optimization problem from becoming distribution independent.
Renyi [113] showed that independence between random variables is equivalent to the condition of vanishing covariance Cov[f (x), g(y)] = 0 for all C 1
functions f, g bounded by L norm 1 on X and Y. In Bach and Jordan
[8], Das and Sen [41], Dauxois and Nkiet [42] and Gretton et al. [58, 59] a
constrained empirical estimate of the above criterion is used. That is, one
studies
(X, Y, F, G) := sup Covemp [f (x), g(y)]
(88)

f,g

subject to f F and g G.

KERNEL METHODS IN MACHINE LEARNING

43

This statistic is often extended to use the entire series 1 , . . . , d of maximal correlations where each of the function pairs (fi , gi ) are orthogonal to
the previous set of terms. More specifically Douxois and Nkiet [42] restrict
F, G to finite-dimensional linear function classes subject to their L2 norm
bounded by 1, Bach and Jordan [8] use functions in the RKHS for which
some sum of the n2 and the RKHS norm on the sample is bounded.
Gretton et al. [58] use functions with bounded RKHS norm only, which
provides necessary and sufficient criteria if kernels are universal. That is,
(X, Y, F, G) = 0 if and only if x and y are independent. Moreover,
tr P Kx P Ky P has the same theoretical properties and it can be computed
much more easily in linear time, as it allows for incomplete Cholesky factorizations. Here Kx and Ky are the kernel matrices on X and Y respectively.
The above criteria can be used to derive algorithms for Independent
Component Analysis (Bach and Jordan [8] and Gretton et al. [58]). While
these algorithms come at a considerable computational cost, they offer very
good performance. For faster algorithms, consider the work of Cardoso [26],
Hyvarinen [73] and Lee et al. [91]. Also, the work of Chen and Bickel [28]
and Yang and Amari [155] is of interest in this context.
Note that a similar approach can be used to develop two-sample tests
based on kernel methods. The basic idea is that for universal kernels the
map between distributions and points on the marginal polytope : p
Exp [(x)] is bijective and, consequently, it imposes a norm on distributions. This builds on the ideas of [52]. The corresponding distance d(p, q) :=
k[p] [q]k leads to a U -statistic which allows one to compute empirical
estimates of distances between distributions efficiently [22].
5.3. Kernel dependency estimation. A large part of the previous discussion revolved around estimating dependencies between samples X and Y for
rather structured spaces Y, in particular, (64). In general, however, such
dependencies can be hard to compute. Weston et al. [153] proposed an algorithm which allows one to extend standard regularized LS regression models,
as described in Section 3.3, to cases where Y has complex structure.
It works by recasting the estimation problem as a linear estimation problem for the map f : (x) (y) and then as a nonlinear pre-image estimation problem for finding y := argminy kf (x) (y)k as the point in Y closest
to f (x).
This problem can be solved directly (Cortes et al. [33]) without the need
for subspace projections. The authors apply it to the analysis of sequence
data.
6. Conclusion. We have summarized some of the advances in the field
of machine learning with positive definite kernels. Due to lack of space,
this article is by no means comprehensive, in particular, we were not able to

44

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

cover statistical learning theory, which is often cited as providing theoretical

support for kernel methods. However, we nevertheless hope that the main
ideas that make kernel methods attractive became clear. In particular, these
include the fact that kernels address the following three major issues of
learning and inference:
They formalize the notion of similarity of data.
They provide a representation of the data in an associated reproducing
kernel Hilbert space.
They characterize the function class used for estimation via the representer theorem [see equations (38) and (86)].
We have explained a number of approaches where kernels are useful. Many
of them involve the substitution of kernels for dot products, thus turning
a linear geometric algorithm into a nonlinear one. This way, one obtains
SVMs from hyperplane classifiers, and kernel PCA from linear PCA. There
is, however, a more recent method of constructing kernel algorithms, where
the starting point is not a linear algorithm, but a linear criterion [e.g., that
two random variables have zero covariance, or that the means of two samples
are identical], which can be turned into a condition involving an efficient
optimization over a large function class using kernels, thus yielding tests
for independence of random variables, or tests for solving the two-sample
problem. We believe that these works, as well as the increasing amount of
work on the use of kernel methods for structured data, illustrate that we
can expect significant further progress in the years to come.
Acknowledgments. We thank Florian Steinke, Matthias Hein, Jakob Macke,
Conrad Sanderson, Tilmann Gneiting and Holger Wendland for comments
and suggestions.
The major part of this paper was written at the Mathematisches
Forschungsinstitut Oberwolfach, whose support is gratefully acknowledged.
National ICT Australia is funded through the Australian Governments
Backing Australias Ability initiative, in part through the Australian Research Council.
REFERENCES
M. and Rozono
[1] Aizerman, M. A., Braverman, E.
er, L. I. (1964). Theoretical
foundations of the potential function method in pattern recognition learning.
Autom. Remote Control 25 821837.
[2] Allwein, E. L., Schapire, R. E. and Singer, Y. (2000). Reducing multiclass
to binary: A unifying approach for margin classifiers. In Proc. 17th International Conf. Machine Learning (P. Langley, ed.) 916. Morgan Kaufmann, San
Francisco, CA. MR1884092

KERNEL METHODS IN MACHINE LEARNING

45

[3] Alon, N., Ben-David, S., Cesa-Bianchi, N. and Haussler, D. (1993). Scalesensitive dimensions, uniform convergence, and learnability. In Proc. of the
34rd Annual Symposium on Foundations of Computer Science 292301. IEEE
Computer Society Press, Los Alamitos, CA. MR1328428
[4] Altun, Y., Hofmann, T. and Smola, A. J. (2004). Gaussian process classification
for segmenting and annotating sequences. In Proc. International Conf. Machine
Learning 2532. ACM Press, New York.
[5] Altun, Y., Smola, A. J. and Hofmann, T. (2004). Exponential families for conditional random fields. In Uncertainty in Artificial Intelligence (UAI) 29. AUAI
Press, Arlington, VA.
[6] Altun, Y., Tsochantaridis, I. and Hofmann, T. (2003). Hidden Markov support
vector machines. In Proc. Intl. Conf. Machine Learning 310. AAAI Press,
Menlo Park, CA.
[7] Aronszajn, N. (1950). Theory of reproducing kernels. Trans. Amer. Math. Soc. 68
337404. MR0051437
[8] Bach, F. R. and Jordan, M. I. (2002). Kernel independent component analysis.
J. Mach. Learn. Res. 3 148. MR1966051
lkopf, B., Smola, A., Taskar, B. and Vish[9] Bakir, G., Hofmann, T., Scho
wanathan, S. V. N. (2007). Predicting Structured Data. MIT Press, Cambridge, MA.
[10] Bamber, D. (1975). The area above the ordinal dominance graph and the area
below the receiver operating characteristic graph. J. Math. Psych. 12 387415.
MR0384214
[11] Barndorff-Nielsen, O. E. (1978). Information and Exponential Families in Statistical Theory. Wiley, New York. MR0489333
[12] Bartlett, P. L. and Mendelson, S. (2002). Rademacher and gaussian complexities: Risk bounds and structural results. J. Mach. Learn. Res. 3 463482.
MR1984026
[13] Basilico, J. and Hofmann, T. (2004). Unifying collaborative and content-based
filtering. In Proc. Intl. Conf. Machine Learning 6572. ACM Press, New York.
[14] Baum, L. E. (1972). An inequality and associated maximization technique in statistical estimation of probabilistic functions of a Markov process. Inequalities
3 18. MR0341782
[15] Ben-David, S., Eiron, N. and Long, P. (2003). On the difficulty of approximately
maximizing agreements. J. Comput. System Sci. 66 496514. MR1981222
[16] Bennett, K. P., Demiriz, A. and Shawe-Taylor, J. (2000). A column generation
algorithm for boosting. In Proc. 17th International Conf. Machine Learning
(P. Langley, ed.) 6572. Morgan Kaufmann, San Francisco, CA.
[17] Bennett, K. P. and Mangasarian, O. L. (1992). Robust linear programming
discrimination of two linearly inseparable sets. Optim. Methods Softw. 1 2334.
[18] Berg, C., Christensen, J. P. R. and Ressel, P. (1984). Harmonic Analysis on
Semigroups. Springer, New York. MR0747302
[19] Bertsimas, D. and Tsitsiklis, J. (1997). Introduction to Linear Programming.
Athena Scientific, Nashua, NH.
[20] Bloomfield, P. and Steiger, W. (1983). Least Absolute Deviations: Theory, Applications and Algorithms. Birkh
auser, Boston. MR0748483
[21] Bochner, S. (1933). Monotone Funktionen, Stieltjessche Integrale und harmonische
Analyse. Math. Ann. 108 378410. MR1512856

46

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

lkopf,
[22] Borgwardt, K. M., Gretton, A., Rasch, M. J., Kriegel, H.-P., Scho
B. and Smola, A. J. (2006). Integrating structured biological data by kernel
maximum mean discrepancy. Bioinformatics (ISMB) 22 e49e57.
[23] Boser, B., Guyon, I. and Vapnik, V. (1992). A training algorithm for optimal margin classifiers. In Proc. Annual Conf. Computational Learning Theory
(D. Haussler, ed.) 144152. ACM Press, Pittsburgh, PA.
[24] Bousquet, O., Boucheron, S. and Lugosi, G. (2005). Theory of classification:
A survey of recent advances. ESAIM Probab. Statist. 9 323375. MR2182250
[25] Burges, C. J. C. (1998). A tutorial on support vector machines for pattern recognition. Data Min. Knowl. Discov. 2 121167.
[26] Cardoso, J.-F. (1998). Blind signal separation: Statistical principles. Proceedings
of the IEEE 90 20092026.
[27] Chapelle, O. and Harchaoui, Z. (2005). A machine learning approach to conjoint
analysis. In Advances in Neural Information Processing Systems 17 (L. K. Saul,
Y. Weiss and L. Bottou, eds.) 257264. MIT Press, Cambridge, MA.
[28] Chen, A. and Bickel, P. (2005). Consistent independent component analysis and
prewhitening. IEEE Trans. Signal Process. 53 36253632. MR2239886
[29] Chen, S., Donoho, D. and Saunders, M. (1999). Atomic decomposition by basis
pursuit. SIAM J. Sci. Comput. 20 3361. MR1639094
[30] Collins, M. (2000). Discriminative reranking for natural language parsing. In Proc.
17th International Conf. Machine Learning (P. Langley, ed.) 175182. Morgan
Kaufmann, San Francisco, CA.
[31] Collins, M. and Duffy, N. (2001). Convolution kernels for natural language.
In Advances in Neural Information Processing Systems 14 (T. G. Dietterich,
S. Becker and Z. Ghahramani, eds.) 625632. MIT Press, Cambridge, MA.
[32] Cook, D., Buja, A. and Cabrera, J. (1993). Projection pursuit indices based
on orthonormal function expansions. J. Comput. Graph. Statist. 2 225250.
MR1272393
[33] Cortes, C., Mohri, M. and Weston, J. (2005). A general regression technique
for learning transductions. In ICML05 : Proceedings of the 22nd International
Conference on Machine Learning 153160. ACM Press, New York.
[34] Cortes, C. and Vapnik, V. (1995). Support vector networks. Machine Learning
20 273297.
[35] Crammer, K. and Singer, Y. (2001). On the algorithmic implementation of multiclass kernel-based vector machines. J. Mach. Learn. Res. 2 265292.
[36] Crammer, K. and Singer, Y. (2005). Loss bounds for online category ranking.
In Proc. Annual Conf. Computational Learning Theory (P. Auer and R. Meir,
eds.) 4862. Springer, Berlin. MR2203253
[37] Cristianini, N. and Shawe-Taylor, J. (2000). An Introduction to Support Vector
Machines. Cambridge Univ. Press.
[38] Cristianini, N., Shawe-Taylor, J., Elisseeff, A. and Kandola, J. (2002). On
kernel-target alignment. In Advances in Neural Information Processing Systems
14 (T. G. Dietterich, S. Becker and Z. Ghahramani, eds.) 367373. MIT Press,
Cambridge, MA.
[39] Culotta, A., Kulp, D. and McCallum, A. (2005). Gene prediction with conditional random fields. Technical Report UM-CS-2005-028, Univ. Massachusetts,
Amherst.
[40] Darroch, J. N. and Ratcliff, D. (1972). Generalized iterative scaling for loglinear models. Ann. Math. Statist. 43 14701480. MR0345337

KERNEL METHODS IN MACHINE LEARNING

47

[41] Das, D. and Sen, P. (1994). Restricted canonical correlations. Linear Algebra Appl.
210 2947. MR1294769
[42] Dauxois, J. and Nkiet, G. M. (1998). Nonlinear canonical analysis and independence tests. Ann. Statist. 26 12541278. MR1647653
[43] Dawid, A. P. (1992). Applications of a general propagation algorithm for probabilistic expert systems. Stat. Comput. 2 2536.
lkopf, B. (2002). Training invariant support vector ma[44] DeCoste, D. and Scho
chines. Machine Learning 46 161190.
[45] Dekel, O., Manning, C. and Singer, Y. (2004). Log-linear models for label
ranking. In Advances in Neural Information Processing Systems 16 (S. Thrun,
L. Saul and B. Sch
olkopf, eds.) 497504. MIT Press, Cambridge, MA.
[46] Della Pietra, S., Della Pietra, V. and Lafferty, J. (1997). Inducing features
of random fields. IEEE Trans. Pattern Anal. Machine Intelligence 19 380393.
[47] Einmal, J. H. J. and Mason, D. M. (1992). Generalized quantile processes. Ann.
Statist. 20 10621078. MR1165606
[48] Elisseeff, A. and Weston, J. (2001). A kernel method for multi-labeled classification. In Advances in Neural Information Processing Systems 14 681687.
MIT Press, Cambridge, MA.
[49] Fiedler, M. (1973). Algebraic connectivity of graphs. Czechoslovak Math. J. 23
298305. MR0318007
[50] FitzGerald, C. H., Micchelli, C. A. and Pinkus, A. (1995). Functions that
preserve families of positive semidefinite matrices. Linear Algebra Appl. 221
83102. MR1331791
[51] Fletcher, R. (1989). Practical Methods of Optimization. Wiley, New York.
MR0955799
[52] Fortet, R. and Mourier, E. (1953). Convergence de la reparation empirique vers

la reparation theorique. Ann. Scient. Ecole

Norm. Sup. 70 266285. MR0061325
[53] Freund, Y. and Schapire, R. E. (1996). Experiments with a new boosting algorithm. In Proceedings of the International Conference on Machine Learing
148146. Morgan Kaufmann, San Francisco, CA.
[54] Friedman, J. H. (1987). Exploratory projection pursuit. J. Amer. Statist. Assoc.
82 249266. MR0883353
[55] Friedman, J. H. and Tukey, J. W. (1974). A projection pursuit algorithm for
exploratory data analysis. IEEE Trans. Comput. C-23 881890.
rtner, T. (2003). A survey of kernels for structured data. SIGKDD Explorations
[56] Ga
5 4958.
[57] Green, P. and Yandell, B. (1985). Semi-parametric generalized linear models.
Proceedings 2nd International GLIM Conference. Lecture Notes in Statist. 32
4455. Springer, New York.
lkopf, B. (2005). Measuring
[58] Gretton, A., Bousquet, O., Smola, A. and Scho
statistical dependence with HilbertSchmidt norms. In Proceedings Algorithmic
Learning Theory (S. Jain, H. U. Simon and E. Tomita, eds.) 6377. Springer,
Berlin. MR2255909
[59] Gretton, A., Smola, A., Bousquet, O., Herbrich, R., Belitski, A., Augath,
lkopf, B. and Logothetis, N. (2005).
M., Murayama, Y., Pauls, J., Scho
Kernel constrained covariance for dependence measurement. In Proceedings
of the Tenth International Workshop on Artificial Intelligence and Statistics
(R. G. Cowell and Z. Ghahramani, eds.) 112119. Society for Artificial Intelligence and Statistics, New Jersey.

48

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

lkopf, B. (2004). A kernel view of the

[60] Ham, J., Lee, D., Mika, S. and Scho
dimensionality reduction of manifolds. In Proceedings of the Twenty-First International Conference on Machine Learning 369376. ACM Press, New York.
[61] Hammersley, J. M. and Clifford, P. E. (1971). Markov fields on finite graphs
and lattices. Unpublished manuscript.
[62] Haussler, D. (1999). Convolutional kernels on discrete structures. Technical Report
UCSC-CRL-99-10, Computer Science Dept., UC Santa Cruz.
lkopf, B. (2005). Maximal margin classification
[63] Hein, M., Bousquet, O. and Scho
for metric spaces. J. Comput. System Sci. 71 333359. MR2168357
[64] Herbrich, R. (2002). Learning Kernel Classifiers: Theory and Algorithms. MIT
Press, Cambridge, MA.
[65] Herbrich, R., Graepel, T. and Obermayer, K. (2000). Large margin rank
boundaries for ordinal regression. In Advances in Large Margin Classifiers
(A. J. Smola, P. L. Bartlett, B. Sch
olkopf and D. Schuurmans, eds.) 115132.
MIT Press, Cambridge, MA. MR1820960
[66] Hettich, R. and Kortanek, K. O. (1993). Semi-infinite programming: Theory,
methods, and applications. SIAM Rev. 35 380429. MR1234637
[67] Hilbert, D. (1904). Grundz
uge einer allgemeinen Theorie der linearen Integralgleichungen. Nachr. Akad. Wiss. G
ottingen Math.-Phys. Kl. II 4991.
[68] Hoerl, A. E. and Kennard, R. W. (1970). Ridge regression: Biased estimation
for nonorthogonal problems. Technometrics 12 5567.
lkopf, B. and Smola, A. J. (2006). A review of kernel meth[69] Hofmann, T., Scho
ods in machine learning. Technical Report 156, Max-Planck-Institut f
ur biologische Kybernetik.
[70] Hotelling, H. (1936). Relations between two sets of variates. Biometrika 28 321
377.
[71] Huber, P. J. (1981). Robust Statistics. Wiley, New York. MR0606374
[72] Huber, P. J. (1985). Projection pursuit. Ann. Statist. 13 435475. MR0790553
rinen, A., Karhunen, J. and Oja, E. (2001). Independent Component Anal[73] Hyva
ysis. Wiley, New York.
[74] Jaakkola, T. S. and Haussler, D. (1999). Probabilistic kernel regression models.
In Proceedings of the 7th International Workshop on AI and Statistics. Morgan
Kaufmann, San Francisco, CA.
[75] Jebara, T. and Kondor, I. (2003). Bhattacharyya and expected likelihood kernels.
Proceedings of the Sixteenth Annual Conference on Computational Learning
Theory (B. Sch
olkopf and M. Warmuth, eds.) 5771. Lecture Notes in Comput.
Sci. 2777. Springer, Heidelberg.
[76] Jensen, F. V., Lauritzen, S. L. and Olesen, K. G. (1990). Bayesian updates in
causal probabilistic networks by local computation. Comput. Statist. Quaterly
4 269282. MR1073446
[77] Joachims, T. (2002). Learning to Classify Text Using Support Vector Machines:
Methods, Theory, and Algorithms. Kluwer Academic, Boston.
[78] Joachims, T. (2005). A support vector method for multivariate performance measures. In Proc. Intl. Conf. Machine Learning 377384. Morgan Kaufmann, San
Francisco, CA.
[79] Jones, M. C. and Sibson, R. (1987). What is projection pursuit? J. Roy. Statist.
Soc. Ser. A 150 136. MR0887823
[80] Jordan, M. I., Bartlett, P. L. and McAuliffe, J. D. (2003). Convexity, classification, and risk bounds. Technical Report 638, Univ. California, Berkeley.

KERNEL METHODS IN MACHINE LEARNING

49

[81] Karush, W. (1939). Minima of functions of several variables with inequalities as

side constraints. Masters thesis, Dept. Mathematics, Univ. Chicago.
[82] Kashima, H., Tsuda, K. and Inokuchi, A. (2003). Marginalized kernels between
labeled graphs. In Proc. Intl. Conf. Machine Learning 321328. Morgan Kaufmann, San Francisco, CA.
[83] Kettenring, J. R. (1971). Canonical analysis of several sets of variables.
Biometrika 58 433451. MR0341750
lkopf, B. (2005). Iterative kernel principal com[84] Kim, K., Franz, M. O. and Scho
ponent analysis for image modeling. IEEE Trans. Pattern Analysis and Machine Intelligence 27 13511366.
[85] Kimeldorf, G. S. and Wahba, G. (1971). Some results on Tchebycheffian spline
functions. J. Math. Anal. Appl. 33 8295. MR0290013
[86] Koltchinskii, V. (2001). Rademacher penalties and structural risk minimization.
IEEE Trans. Inform. Theory 47 19021914. MR1842526
[87] Kondor, I. R. and Lafferty, J. D. (2002). Diffusion kernels on graphs and other
discrete structures. In Proc. International Conf. Machine Learning 315322.
Morgan Kaufmann, San Francisco, CA.
[88] Kuhn, H. W. and Tucker, A. W. (1951). Nonlinear programming. Proc. 2nd
Berkeley Symposium on Mathematical Statistics and Probabilistics 481492.
Univ. California Press, Berkeley. MR0047303
[89] Lafferty, J., Zhu, X. and Liu, Y. (2004). Kernel conditional random fields: Representation and clique selection. In Proc. International Conf. Machine Learning
21 64. Morgan Kaufmann, San Francisco, CA.
[90] Lafferty, J. D., McCallum, A. and Pereira, F. (2001). Conditional random
fields: Probabilistic modeling for segmenting and labeling sequence data. In
Proc. International Conf. Machine Learning 18 282289. Morgan Kaufmann,
San Francisco, CA.
[91] Lee, T.-W., Girolami, M., Bell, A. and Sejnowski, T. (2000). A unifying
framework for independent component analysis. Comput. Math. Appl. 39 1
21. MR1766376
[92] Leslie, C., Eskin, E. and Noble, W. S. (2002). The spectrum kernel: A string
kernel for SVM protein classification. In Proceedings of the Pacific Symposium
on Biocomputing 564575. World Scientific Publishing, Singapore.
[93] Lo`
eve, M. (1978). Probability Theory II, 4th ed. Springer, New York. MR0651018
[94] Magerman, D. M. (1996). Learning grammatical structure using statistical
decision-trees. Proceedings ICGI. Lecture Notes in Artificial Intelligence 1147
121. Springer, Berlin.
[95] Mangasarian, O. L. (1965). Linear and nonlinear separation of patterns by linear
programming. Oper. Res. 13 444452. MR0192918
[96] McCallum, A., Bellare, K. and Pereira, F. (2005). A conditional random field
for discriminatively-trained finite-state string edit distance. In Conference on
Uncertainty in AI (UAI) 388. AUAI Press, Arlington, VA.
[97] McCullagh, P. and Nelder, J. A. (1983). Generalized Linear Models. Chapman
and Hall, London. MR0727836
[98] Mendelson, S. (2003). A few notes on statistical learning theory. Advanced Lectures
on Machine Learning (S. Mendelson and A. J. Smola, eds.). Lecture Notes in
Artificial Intelligence 2600 140. Springer, Heidelberg.
[99] Mercer, J. (1909). Functions of positive and negative type and their connection
with the theory of integral equations. Philos. Trans. R. Soc. Lond. Ser. A Math.
Phys. Eng. Sci. A 209 415446.

50

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

tsch, G., Weston, J., Scho

lkopf, B., Smola, A. J. and
[100] Mika, S., Ra
ller, K.-R. (2003). Learning discriminative and invariant nonlinear feaMu
tures. IEEE Trans. Pattern Analysis and Machine Intelligence 25 623628.
[101] Minsky, M. and Papert, S. (1969). Perceptrons: An Introduction to Computational
Geometry. MIT Press, Cambridge, MA.
[102] Morozov, V. A. (1984). Methods for Solving Incorrectly Posed Problems. Springer,
New York. MR0766231
[103] Murray, M. K. and Rice, J. W. (1993). Differential Geometry and Statistics.
Chapman and Hall, London. MR1293124
lkopf, B. and Smola, A. J. (2000). Natural regularization in
[104] Oliver, N., Scho
SVMs. In Advances in Large Margin Classifiers (A. J. Smola, P. L. Bartlett,
B. Sch
olkopf and D. Schuurmans, eds.) 5160. MIT Press, Cambridge, MA.
MR1820960
[105] OSullivan, F., Yandell, B. and Raynor, W. (1986). Automatic smoothing of
regression functions in generalized linear models. J. Amer. Statist. Assoc. 81
96103. MR0830570
[106] Parzen, E. (1970). Statistical inference on time series by RKHS methods. In Proceedings 12th Biennial Seminar (R. Pyke, ed.) 137. Canadian Mathematical
Congress, Montreal. MR0275616
[107] Platt, J. (1999). Fast training of support vector machines using sequential minimal optimization. In Advances in Kernel MethodsSupport Vector Learning
(B. Sch
olkopf, C. J. C. Burges and A. J. Smola, eds.) 185208. MIT Press,
Cambridge, MA.
[108] Poggio, T. (1975). On optimal nonlinear associative recall. Biological Cybernetics
19 201209. MR0503978
[109] Poggio, T. and Girosi, F. (1990). Networks for approximation and learning. Proceedings of the IEEE 78 14811497.
[110] Press, W. H., Teukolsky, S. A., Vetterling, W. T. and Flannery, B. P.
(1994). Numerical Recipes in C. The Art of Scientific Computation. Cambridge
Univ. Press. MR1880993
[111] Rasmussen, C. E. and Williams, C. K. I. (2006). Gaussian Processes for Machine
Learning. MIT Press, Cambridge, MA.
tsch, G., Sonnenburg, S., Srinivasan, J., Witte, H., Mu
ller, K.-R., Som[112] Ra
lkopf, B. (2007). Improving the Caenorhabditis elegans
mer, R. J. and Scho
genome annotation using machine learning. PLoS Computational Biology 3 e20
doi:10.1371/journal.pcbi.0030020.
[113] R
enyi, A. (1959). On measures of dependence. Acta Math. Acad. Sci. Hungar. 10
441451. MR0115203
[114] Rockafellar, R. T. (1970). Convex Analysis. Princeton Univ. Press. MR0274683
[115] Schoenberg, I. J. (1938). Metric spaces and completely monotone functions. Ann.
Math. 39 811841. MR1503439
lkopf, B. (1997). Support Vector Learning. R. Oldenbourg Verlag, Munich.
[116] Scho
Available at http://www.kernel-machines.org.
lkopf, B., Platt, J., Shawe-Taylor, J., Smola, A. J. and Williamson,
[117] Scho
R. C. (2001). Estimating the support of a high-dimensional distribution. Neural
Comput. 13 14431471.
lkopf, B. and Smola, A. (2002). Learning with Kernels. MIT Press, Cam[118] Scho
bridge, MA.
lkopf, B., Smola, A. J. and Mu
ller, K.-R. (1998). Nonlinear component
[119] Scho
analysis as a kernel eigenvalue problem. Neural Comput. 10 12991319.

KERNEL METHODS IN MACHINE LEARNING

51

lkopf, B., Smola, A. J., Williamson, R. C. and Bartlett, P. L. (2000).

[120] Scho
New support vector algorithms. Neural Comput. 12 12071245.
lkopf, B., Tsuda, K. and Vert, J.-P. (2004). Kernel Methods in Computa[121] Scho
tional Biology. MIT Press, Cambridge, MA.
[122] Sha, F. and Pereira, F. (2003). Shallow parsing with conditional random fields.
In Proceedings of HLT-NAACL 213220. Association for Computational Linguistics, Edmonton, Canada.
[123] Shawe-Taylor, J. and Cristianini, N. (2004). Kernel Methods for Pattern Analysis. Cambridge Univ. Press.
lkopf, B. and Schuurmans, D. (2000).
[124] Smola, A. J., Bartlett, P. L., Scho
Advances in Large Margin Classifiers. MIT Press, Cambridge, MA. MR1820960
[125] Smola, A. J. and Kondor, I. R. (2003). Kernels and regularization on graphs.
Proc. Annual Conf. Computational Learning Theory (B. Sch
olkopf and M. K.
Warmuth, eds.). Lecture Notes in Comput. Sci. 2726 144158. Springer, Heidelberg.
lkopf, B. (1998). On a kernel-based method for pattern
[126] Smola, A. J. and Scho
recognition, regression, approximation and operator inversion. Algorithmica 22
211231. MR1637511
lkopf, B. and Mu
ller, K.-R. (1998). The connection between
[127] Smola, A. J., Scho
regularization operators and support vector kernels. Neural Networks 11 637
649.
[128] Steinwart, I. (2002). On the influence of the kernel on the consistency of support
vector machines. J. Mach. Learn. Res. 2 6793. MR1883281
[129] Steinwart, I. (2002). Support vector machines are universally consistent. J. Complexity 18 768791. MR1928806
[130] Stewart, J. (1976). Positive definite functions and generalizations, an historical
survey. Rocky Mountain J. Math. 6 409434. MR0430674
[131] Stitson, M., Gammerman, A., Vapnik, V., Vovk, V., Watkins, C. and Weston, J. (1999). Support vector regression with ANOVA decomposition kernels. In Advances in Kernel MethodsSupport Vector Learning (B. Sch
olkopf,
C. J. C. Burges and A. J. Smola, eds.) 285292. MIT Press, Cambridge, MA.
[132] Taskar, B., Guestrin, C. and Koller, D. (2004). Max-margin Markov networks.
In Advances in Neural Information Processing Systems 16 (S. Thrun, L. Saul
and B. Sch
olkopf, eds.) 2532. MIT Press, Cambridge, MA.
[133] Taskar, B., Klein, D., Collins, M., Koller, D. and Manning, C. (2004). Maxmargin parsing. In Empirical Methods in Natural Language Processing 18.
Association for Computational Linguistics, Barcelona, Spain.
[134] Tax, D. M. J. and Duin, R. P. W. (1999). Data domain description by support
vectors. In Proceedings ESANN (M. Verleysen, ed.) 251256. D Facto, Brussels.
[135] Tibshirani, R. (1996). Regression shrinkage and selection via the lasso. J. R. Stat.
Soc. Ser. B Stat. Methodol. 58 267288. MR1379242
[136] Tikhonov, A. N. (1963). Solution of incorrectly formulated problems and the regularization method. Soviet Math. Dokl. 4 10351038.
[137] Tsochantaridis, I., Joachims, T., Hofmann, T. and Altun, Y. (2005). Large
margin methods for structured and interdependent output variables. J. Mach.
Learn. Res. 6 14531484. MR2249862
[138] van Rijsbergen, C. (1979). Information Retrieval, 2nd ed. Butterworths, London.
[139] Vapnik, V. (1982). Estimation of Dependences Based on Empirical Data. Springer,
Berlin. MR0672244

52

T. HOFMANN, B. SCHOLKOPF
AND A. J. SMOLA

[140] Vapnik, V. (1995). The Nature of Statistical Learning Theory. Springer, New York.
MR1367965
[141] Vapnik, V. (1998). Statistical Learning Theory. Wiley, New York. MR1641250
[142] Vapnik, V. and Chervonenkis, A. (1971). On the uniform convergence of relative
frequencies of events to their probabilities. Theory Probab. Appl. 16 264281.
[143] Vapnik, V. and Chervonenkis, A. (1991). The necessary and sufficient conditions
for consistency in the empirical risk minimization method. Pattern Recognition
and Image Analysis 1 283305.
[144] Vapnik, V., Golowich, S. and Smola, A. J. (1997). Support vector method for
function approximation, regression estimation, and signal processing. In Advances in Neural Information Processing Systems 9 (M. C. Mozer, M. I. Jordan
and T. Petsche, eds.) 281287. MIT Press, Cambridge, MA.
[145] Vapnik, V. and Lerner, A. (1963). Pattern recognition using generalized portrait
method. Autom. Remote Control 24 774780.
[146] Vishwanathan, S. V. N. and Smola, A. J. (2004). Fast kernels for string and
tree matching. In Kernel Methods in Computational Biology (B. Sch
olkopf,
K. Tsuda and J. P. Vert, eds.) 113130. MIT Press, Cambridge, MA.
[147] Vishwanathan, S. V. N., Smola, A. J. and Vidal, R. (2007). BinetCauchy
kernels on dynamical systems and its application to the analysis of dynamic
scenes. Internat. J. Computer Vision 73 95119.
[148] Wahba, G. (1990). Spline Models for Observational Data. SIAM, Philadelphia.
MR1045442
[149] Wahba, G., Wang, Y., Gu, C., Klein, R. and Klein, B. (1995). Smoothing spline
ANOVA for exponential families, with application to the Wisconsin Epidemiological Study of Diabetic Retinopathy. Ann. Statist. 23 18651895. MR1389856
[150] Wainwright, M. J. and Jordan, M. I. (2003). Graphical models, exponential
families, and variational inference. Technical Report 649, Dept. Statistics, Univ.
California, Berkeley.
[151] Watkins, C. (2000). Dynamic alignment kernels. In Advances in Large Margin
Classifiers (A. J. Smola, P. L. Bartlett, B. Sch
olkopf and D. Schuurmans, eds.)
3950. MIT Press, Cambridge, MA. MR1820960
[152] Wendland, H. (2005). Scattered Data Approximation. Cambridge Univ. Press.
MR2131724
lkopf, B. and Vapnik, V.
[153] Weston, J., Chapelle, O., Elisseeff, A., Scho
(2003). Kernel dependency estimation. In Advances in Neural Information Processing Systems 15 (S. T. S. Becker and K. Obermayer, eds.) 873880. MIT
Press, Cambridge, MA. MR1820960
[154] Whittaker, J. (1990). Graphical Models in Applied Multivariate Statistics. Wiley,
New York. MR1112133
[155] Yang, H. H. and Amari, S.-I. (1997). Adaptive on-line learning algorithms for
blind separationmaximum entropy and minimum mutual information. Neural
Comput. 9 14571482.
[156] Zettlemoyer, L. S. and Collins, M. (2005). Learning to map sentences to logical form: Structured classification with probabilistic categorial grammars. In
Uncertainty in Artificial Intelligence UAI 658666. AUAI Press, Arlington,
Virginia.
tsch, G., Mika, S., Scho
lkopf, B., Lengauer, T. and Mu
ller,
[157] Zien, A., Ra
K.-R. (2000). Engineering support vector machine kernels that recognize translation initiation sites. Bioinformatics 16 799807.

KERNEL METHODS IN MACHINE LEARNING

T. Hofmann
Darmstadt University of Technology
Department of Computer Science
Darmstadt
Germany
E-mail: hofmann@int.tu-darmstadt.de

53

lkopf
B. Scho
Max Planck Institute
for Biological Cybernetics
bingen
Tu
Germany
E-mail: bs@tuebingen.mpg.de

A. J. Smola
Statistical Machine Learning Program
National ICT Australia
Canberra
Australia
E-mail: Alex.Smola@nicta.com.au