Problem Set #1. Due Sept. 9 2020.: MAE 501 - Fall 2020. Luc Deike, Anastasia Bizyaeva, Jiarong Wu September 2, 2020

Problem Set #1. Due Sept. 9 2020.
MAE 501 - Fall 2020. Luc Deike, Anastasia Bizyaeva, Jiarong Wu

September 2, 2020
Homework is due by 09/09/2020 midnight ET. We will correct anonymously, so please

do not indicate your name when you scan and submit your answer via gradescope.
Exercise 1. Expansion and basis.
a) Expand x = (3, 1 − i, 0) by any means in terms of
(i) e1 = (1, 0, 0), e2 = (1, 1, 0), e3 = (1, 1, 1)
(ii) e1 = (1, 0, 0), e2 = (0, 1, 0), e3 = (0, 0, 1)
(iii) e1 = (i, 0, 0), e2 = (0, i, 1), e3 = (0, 0, i)
b) Verify that the following set is linearly independent

e1 = (i, 0, 0), e2 = (0, 1, 0), e3 = (1, 0, 1)
c) Show that the following sets are bases for R2

(i) e1 = (1, 1), e2 = (1, −1)
(ii) e1 = (1, 0), e2 = (1, 1).
Exercise 2. Orthogonalizing vectors.
Suppose that a and b are any n-vectors. Show that we can always find a scalar γ so that
(a − γb) is orthogonal to b, and that γ is unique if b is different than the null vector 0. Give a
formula for the scalar γ. In other words, we can always subtract a multiple of a vector from
another one, so that the result is orthogonal to the original vector. The orthogonalization
step in the Gram–Schmidt algorithm is an application of this.
Exercise 3. Function spaces
a) We consider the function space C(a, b) of all continuous functions x(t) defined over
a ≤ t ≤ b. Scalar multiplication and function addition is defined in the the obvious way.
Verify that this is a linear vector space called a function space. Note that the function space
C(a, b) is infinite dimensional.
b) We define the inner product

Z b
(x, y) = x(t)y(t)dt (1)
a
Verify that it fulfills the definition of an inner product.
1
c) We remark that C(a, b) is not complete. Indeed you can define a sequence xn (t)
such that xn (t) = 0 for −1 ≤ t ≤ 0; xn (t) = nt for 0 ≤ t ≤ 1/n; and xn (t) = 1 for
1/n ≤ t ≤ 1, which is in C(−1, 1). This sequence xn (t) converges to the Heaviside function
H(t) as n → ∞, but H(t) is not part of C(−1, 1), since it is not continuous. We say that the
space does not contain all its limit points, it is not complete. This situation is unfortunate,
as complete spaces (a space that contains all its limit points) are more tractable. We can
see such spaces as having holes in it. If we can plug all the holes, then the space is complete
and the resulting space is called a Lesbegue space, L2 (a, b), namely the space of all functions
Rb
x(t) that are square-integrable over the interval, that is such that a |x(t)|2 dt < ∞, i.e. the
vector x(t) must be of finite length.
c.i) Is x(t) = 6t2 in L2 (−1, 1)? What about x(t) = 1/t?
c.ii) Verify that the following set is linearly independent: e1 = t, e2 = sin(t), e3 = cos(t)
in L2 (0, π).
d) Lets build a basis for L2 (−1, 1) as follows. We define φ1 (t) = 1. Next, seek φ2 (t) =
a1 + bt. Determine a and b such that the two vectors are orthogonal, (φ1 , φ2 ) = 0. Similarly,
find φ3 (t). This procedure is called Gram-Schmidt orthogonalization. Normalize each vector
by its length and obtain an orthonormal basis φj (t). Alternatively, scale each vector so that
it is unity at t = 1 and to obtain the Legendre polynomials, Pj (t).
e) The set of orthonormal vectors φj (t), or Pj (t) are basis for L2 (−1, 1). That is, we can
expand any x(t) in L2 (−t, t) as X
x(t) = αj φj (t) (2)
j
Determine the coefficient αj ’s. Hint, you should compute the inner product (x(t), φk (t)).
f) Expand x(t) = |t| over the interval [−1, 1] in terms of the φj (t). Make a graph of the
expansion and compare with the exact solution. Perform the expansion in terms of the Pj
and verify that you obtain the same result.
g) Another basis. We will see later that the functions 1, cos(πt/l), cos(2πt/l), ...,
sin(πt/l), sin(2πt/l), ..., form an orthogonal basis for L2 (−l, l). Expanding a given x(t),
∞
X
x(t) = [aj cos(jπt/l) + bj sin(jπt/l)] (3)
j=0
Compute bk by applying (x(t), sin(kπt/l)) and ak in a similar way to obtain the Fourier series
decomposition.
Exercise 4. Matrix manipulation.
Given the matrices A, B, x,  
2 0
 1 4
A= 
1 − i 3 
0 −1
2

3 2
B=
0 1

3
x=
i
work out whichever of the products AB, BA, Ax, xA, Bx, xB, AA, BB, xx are defined and
compute them if they are defined.
Exercise 5. Matrix manipulation.

Given that A ∈ Rm×n , show via manipulation using index notation that AT A ∈ Rn×n
is symmetric. Given that A ∈ Rm×n , show via manipulation using index notation that
AAT ∈ Rm×m is symmetric.
Exercise 6. Least square solution for an over-determined system.
Ax = b, A ∈ Rm×n ; x ∈ Rn ; b ∈ Rm , (4)
Find the least-squares solution for an over-determined system given by equation (4) with
m > n: In general, there is no exact solution for such systems. The residual error in any
solution x is given by q = Ax − b, and the best we can hope for is to minimize the residual
error, ||q|| or ||q||2 .
Consider the least-squares method where we seek to minimize Q (x) = 12 q T q. To minimize
∂Q
Q (x), we demand ∂x i
= 0, i = 1, ..., n.
Express Q(x) in index notation and perform the indicated differentiation to show that
the least-squares solution must satisfy
AT Ax = AT b, (5)
(You dealt with AT A in an exercise in Exercise 5.) Equation (5), referred to as normal
equations, can be solved if AT A has full rank (and hence invertible).
Exercise 7. A surprising discovery. An intern at a quantitative hedge fund examines the

daily returns of around 400 stocks over one year (which has 250 trading days). She tells her
supervisor that she has discovered that the returns of one of the stocks, Google (GOOG),
can be expressed as a linear combination of the others, which include many stocks that are
unrelated to Google (say, in a different type of business or sector). Her supervisor then
says: “It is overwhelmingly unlikely that a linear combination of the returns of unrelated
companies can reproduce the daily return of GOOG. So you’ve made a mistake in your
calculations.” Is the supervisor right? Did the intern make a mistake? Give a very brief
explanation. Think about the definition of linear independence and linear dependence and
the dimensions of the vector space and vectors.

Problem Set #1. Due Sept. 9 2020.: MAE 501 - Fall 2020. Luc Deike, Anastasia Bizyaeva, Jiarong Wu September 2, 2020

Uploaded by

Copyright:

Available Formats

Problem Set #1. Due Sept. 9 2020.: MAE 501 - Fall 2020. Luc Deike, Anastasia Bizyaeva, Jiarong Wu September 2, 2020

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Problem Set #1. Due Sept. 9 2020.: MAE 501 - Fall 2020. Luc Deike, Anastasia Bizyaeva, Jiarong Wu September 2, 2020

Uploaded by

Copyright:

Available Formats

Problem Set #1. Due Sept. 9 2020.

MAE 501 - Fall 2020. Luc Deike, Anastasia Bizyaeva, Jiarong Wu

Homework is due by 09/09/2020 midnight ET. We will correct anonymously, so please

b) Verify that the following set is linearly independent

c) Show that the following sets are bases for R2

b) We define the inner product

Verify that it fulfills the definition of an inner product.

Exercise 5. Matrix manipulation.

Exercise 6. Least square solution for an over-determined system.

Exercise 7. A surprising discovery. An intern at a quantitative hedge fund examines the

You might also like