Scribd
Upload
14 views

Notes-2-Supervised-Hebb Learning

Uploaded by

kapoornavodit
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
14 views

Notes-2-Supervised-Hebb Learning

Uploaded by

kapoornavodit
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Hebb Learning

Chapter – 6,7, Textbook


Introduction:

A neural network is trained with the help of a set of patterns known as the training vectors. The outputs
for these vectors might be, or might not be, known beforehand. When these are known, and that
knowledge is employed in the training process, the training is termed as supervised learning. Otherwise,
the learning is said to be unsupervised.

Heb Net:
A single-layer feedforward neural net trained through the Hebb learning rule is known as a Hebb Net.

Hebb Rule:
The Hebb rule is one of the earliest learning rules for ANNs. According to this rule the weight adjustment
is computed as

where t is the target activation.

There are certain points to be kept in mind regarding the Hebb learning rule. First, Hebb rule cannot
learn when the target is 0. This is because the weight adjustment Δwi becomes zero when t = 0,
irrespective of the value of xi. Hence, obviously, the Hebb rule results in better learning if the input /
output both are in bipolar form. Th e most striking limitation of the Hebb rule is it does not guarantee to
learn a classification instance even if the classes are linearly separable.

Algorithm for Heb Learning:

It should be noted that the input unit X0 and the associated weight w0 play the role of the bias. The
activation of X0 is always kept at 1. Hence, the expression for adjustment of w0 becomes
Hebb Learning
Chapter – 6,7, Textbook
Example for Heb Learning:

Realizing the logical AND function through Hebb learning

To realize a two input AND function we need a net with two input units and one output unit. A bias is also
needed. Hence the structure of the required neural net should be as shown in the following figure.

Moreover, the input and output signals must be in bipolar form, rather than the binary form, so that the
net may be trained properly. Considering the truth table of AND operation, and the fact that the bias is
permanently set to 1, we get the training set depicted in the following Table.

During the training process, all weights are initialized to 0. Therefore initially w0 = w1 = w2 = 0. At each
training instance, the weights are changed according to the formula
wi (new) = wi (old) + Δwi
where Δwi, the increment in wi, is computed as Δwi = xi.× t.
After initialization, the progress of the learning process by the network is shown in following Table.

You might also like