6.

034f Neural Net Notes

October 28, 2010
These notes are a supplement to material presented in lecture. I lay out the mathematics more prettily
and extend the analysis to handle multiple-neurons per layer. Also, I develop the back propagation
rule, which is often needed on quizzes.
I use a notation that I think improves on previous explanations. The reason is that the notation
here plainly associates each input, output, and weight with a readily identified neuron, a left-side
one and a right-side one. When you arrive at the update formulas, you will have less trouble relating
the variables in the formulas to the variables in a diagram.
One the other hand, seeing yet another notation may confuse you, so if you already feel com­
fortable with a set of update formulas, you will not gain by reading these notes.

The sigmoid function

The sigmoid function, y = 1/(1 + e−x ), is used instead of a step function in artificial neural nets
because the sigmoid is continuous, whereas a step function is not, and you need continuity whenever
you want to use gradient ascent. Also, the sigmoid function has several desirable qualities. For
example, the sigmoid function’s value, y, approaches 1 as x becomes highly positive; 0 as x becomes
highly negative; and equals 1/2 when x = 0.
Better yet, the sigmoid function features a remarkably simple derivative of the output, y, with
respect to the input, x:

dy d 1
= ( )
dx dx 1 + e−x
d
= (1 + e−x )−1
dx
= − 1 × (1 + e−x )−2 × e−x × −1
1 e−x
= ×
1 + e−x 1 + e−x
1 1 + e−x − 1
= ×
1 + e−x 1 + e−x
1 1 + e−x 1
= × ( − )
1 + e−x 1 + e−x 1 + e−x
=y(1 − y)
Thus, remarkably, the derivative of the output with respect to the input is expressed as a simple
function of the output.

The performance function

The standard performance function for gauging how well a neural net is doing is given by the
following:

1
P = − (dsample − osample )2
2
 2

where P is the performance function, dsample is the desired output for some specific sample and
osample is the observed output for that sample. From this point forward, assume that d and o are
the desired and observed outputs for a specific sample so that we need not drag a subscript around
as we work through the algebra.
The reason for choosing the given formula for P is that the formula has convenient properties.
The formula yields a maximum at o = d and monotonically decreases as o deviates from d . Moreover,
the derivative of P with respect to o is simple:

dP d 1
= [− (d − o)2 ]
do do 2
2
= − × (d − o)1 × −1
2
=d −o

Gradient ascent
Backpropagation is a specialization of the idea of gradient ascent. You are trying to find the maximum
of a performance function P, by changing the weights associated with neurons, so you move in the
direction of the gradient in a space that gives P as a function of the weights, w. That is, you move in
the direction of most rapid ascent if we take a step in the direction with components governed by the
following formula, which shows how much to change a weight, w, in terms of a partial derivative:

∂P
Δw ∝
∂w
The actual change is influenced by a rate constant, α; accordingly, the new weight, w , is given by
the following:

∂P
w = w + α ×
∂w

Gradient descent
If the performance function were 12 (dsample − osample )2 instead of − 12 (dsample − osample )2 , then
you would be searching for the minimum rather than the maximum of P, and the change in w would
be subtracted from w instead of added, so w would be w − α × ∂∂wP instead of w + α × ∂∂wP . The two
sign changes, one in the performance function and the other in the update formula cancel, so in the
end, you get the same result whether you use gradient ascent, as I prefer, or gradient descent.

The simplest neural net

Consider the simplest possible neural net: one input, one output, and two neurons, the left neuron
and the right neuron. A net with two neurons is the smallest that illustrates how the derivatives can
be computed layer by layer.
 3

Wl Wr

il pl ol ir pr or
x Sigmoid x Sigmoid

Left neuron Right neuron

Note that the subscripts indicate layer. Thus, il , wl , pl , and ol are the input, weight, product, and
output associated with the neuron on the left while ir , wr , pr , and or are the input, weight, product,
and output associated with the neuron on the right. Of course, ol = ir .
Suppose that the output of the right neuron, or , is the value that determines performance P. To
compute the partial derivative of P with respect to the weight in the right neuron, wr , you need the
chain rule, which allows you to compute partial derivatives of one variable with respect to another
in terms of an intermediate variable. In particular, for wr , you have the following, taking or to be
the intermediate variable:

∂P ∂P ∂or
= ×
∂wr ∂or ∂wr
∂ or ∂ or ∂ pr
Now, you can repeat, using the chain-rule to turn ∂ wr into ∂ pr × ∂ wr :

∂P ∂P ∂or ∂pr
= × ×
∂wr ∂or ∂pr ∂wr
∂ pr ∂(wr ×ol )
Conveniently, you have seen two of the derivatives already, and the third, ∂ wr = ∂ wr , is easy to
compute:

∂P
= [(d − or )] × [or (1 − or )] × [ir ]
∂wr

Repeating the analysis for wl yields the following. Each line is the same as the previously, except
that one more partial derivative is expanded using the chain rule:

∂P ∂P ∂or
= ×
∂wl ∂or ∂wl
∂P ∂or ∂pr
= × ×
∂or ∂pr ∂wl
∂P ∂or ∂pr ∂ol
= × × ×
∂or ∂pr ∂ol ∂wl
∂P ∂or ∂pr ∂ol ∂pl
= × × × ×
∂or ∂pr ∂ol ∂pl ∂wl
=[(d − or )] × [or (1 − or )] × [wr ] × [ol (1 − ol )] × [il ]
 4

Thus, the derivative consists of products of terms that have already been computed and terms in the
vicinity of wl . This is clearer if you write the two derivatives next to one another:

∂P
=(d − or ) × or (1 − or ) × ir
∂wr

∂P

=(d − or ) × or (1 − or ) × wr × ol (1 − ol ) × il
∂wl
You can simplify the equations by defining δs as follows, where each delta is associated with either
the left or right neuron:

δr =or (1 − or ) × (d − or )
δl =ol (1 − ol ) × wr × δr
Then, you can write the partial derivatives with the δs:

∂P
=ir × δr
∂wr
∂P
=il × δl
∂wl
If you add more layers to the front of the network, each weight has a partial derivatives that
is computed like the partial derivative of the weight of the left neuron. That is, each has a partial
derivative determined by its input and its delta, where its delta in turn is determined by its output,
the weight to its right, and the delta to its right. Thus, for the weights in the final layer, you compute
the change as follows, where I use f as the subscript instead of r to emphasize that the computation
is for the neuron in the final layer:

Δwf = α × if × δf
where

δf = of (1 − of ) × (d − of )

For all other layers, you compute the change as follows:

Δwl = α × il × δl
where

δl = ol (1 − ol ) × wr × δr

More neurons per layers

Of course, you really want back propagation formulas for not only any number of layers but also for
any number of neurons per layer, each of which can have multiple inputs, each with its own weight.
Accordingly, you need to generalize in another direction, allowing multiple neurons in each layer
and multiple weights attached to each neuron.
The generalization is an adventure in summations, with lots of subscripts to keep straight, but
in the end, the result matches intuition. For the final layer, there may be many neurons, so the
formula’s need an index, k, indicating which final node neuron is in play. For any weight contained
 5

in the final-layer neuron, fk , you compute the change as follows from the input corresponding to the
weight and from the δ associated with the neuron:

Δw =α × i × δfk
δfk =ofk (1 − ofk ) × (dk − ofk )

Note that the output of each final-layer neuron output is subtracted from the output desired for that
neuron.

For other layers, there may also be many neurons, and the output of each may influence all the
neurons in the next layer to the right. The change in weight has to account for what happens to all
of those neurons to the right, so a summation appears, but otherwise you compute the change, as
before, from the input corresponding to the weight and from the δ associated with the neuron:

Δw =α × i × δli
�
δli =oli (1 − oli ) × wli →rj × δrj
j

Note that wli →rj is the weight that connects the jth right-side neuron to the output of the ith left-side
neuron.

Summary

Once you understood how to derive the formulas, you can combine and simplify them in preparation
for solving problems. For each weight, you compute the weight’s change from the input correspond­
ing to the weight and from the δ associated with the neuron. Assuming that δ is the delta associated
with that neuron, you have the following, where w→rj is the weight connecting the output of the
neuron you are working on, the ith left-side neuron, to the jth right-side neuron, and δ is the δ
rj
associated with that right-side neuron.

δo =o(1 − o) × (d − o) for the final layer

�
δli =oli (1 − oli ) × wli →rj × δrj otherwise
j

That is, you computed change in a neuron’s w, in every layer, by multiplying α times the
neuron’s input times its δ. The δ is determined for all but the final layer in terms of the neuron’s
output and all the weights that connect that output to neurons in the layer to the right and the δs
associated with those right-side neurons. The δ for each neuron in the final layer is determined only
by the output of that neuron and by the difference between the desired output and the actual output
of that neuron.
 6

Weights and deltas in layer to the right

Neuron with weight to be adjusted w→r

1
w Weight to be adjusted x

x
Σ ∫

i x
δ1
x
o
x
Σ ∫
x
w→r
N
x

x
Σ ∫
x δΝ
MIT OpenCourseWare
http://ocw.mit.edu

6.034 Artificial Intelligence

Fall 2010

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.