You are on page 1of 35

Artificial Neural Networks

The Brain
The human brain contains about
10 billion nerve cells, or neurons.
On average, each neuron is connected
to other neurons through about
10 000 synapses.
Synapses mediates the interaction between neurons (60 trillion)
Highly complex
Nonlinear
Parallel Computer
Organize neurons to perform computations such as
Pattern Recognition

2
The basic computational unit in the nervous
system is the nerve cell, or neuron. A neuron
has:

Dendrites (inputs)
Cell body
Axon (output)
Neurons
A neuron receives input from other neurons (typically many thousands). Inputs
sum (approximately). Once input exceeds a critical level, the neuron discharges a
spike - an electrical pulse that travels from the body, down the axon, to the next
neuron(s) (or other receptors). This spiking event is also called depolarization, and
is followed by a refractory period, during which the neuron is unable to fire.

The axon endings (Output Zone) almost touch the dendrites or cell body of the
next neuron. Transmission of an electrical signal from one neuron to the next is
effected by neurotransmitters, chemicals which are released from the first neuron
and which bind to receptors in the second. This link is called a synapse. The extent
to which the signal from one neuron is passed on to the next depends on many
factors, e.g. the amount of neurotransmitter available, the number and arrangement
of receptors, amount of neurotransmitter reabsorbed, etc.
3
Artificial Neural Network

Inspired from its inception by the human brain
NN is designed to model the way in which the brain performs
a particular task.
Powerful technique to solve many real world problems.
Ability to learn from experience in order to improve their
performance
Adapt themselves to changes in the environment.
Able to deal with incomplete information or noisy data.
very effective especially in situations where it is not possible
to define the rules or steps that lead to the solution of a
problem.

Definition
A neural Network is a massively parallel distributed processor
made up of simple processing units, which can store
experimental knowledge and make it available for use.




4
Artificial Neuron
ANNs incorporate the two fundamental
components of biological neural nets:
Neurons as Nodes
Synapses as Weights
Neuron Node
5
Synapse Weight
Neuron Model
Three basic element are
Synapses or connecting link
characterized by a weight or
strength.
Adder for summing the input signals.
Activation Function (Squashing Function) for limiting the
amplitude of the o/p
x
1
, x
2
x
m
are the inputs and w
k1
w
k2,
..w
km
are the respective
weights.
b
k
is the externally applied bias which increases or decrease the
net input of the activation function f(x)
Neuron k :
u
k
= w
kj
x
j
j=1.m and y
k
= f(u
k
+ b
k
) v
k
= u
k
+ b
k




6
W
km
b
k
f(x)
x
1
x
2
x
m
W
k1
W
k2
v
k
y
k
Activation Function
Threshold Function



Piecewise Linear Function



Sigmoid Function
7
Definition of Learning
Learning is a process by which the free parameters of a neural
network are adapted through a process of stimulation by the
environment in which the network is embedded. The type of the
learning is determined by the manner in which the parameter
changes take place.

The ability of the neural network (NN) to learn from its environment
and to improve its performance through learning.
- The NN is stimulated by an environment
- The NN undergoes changes in its free parameteres
- The NN responds in a new way to the environment

8
Five Basic Learning Rules

Error-correction learning <- optimum filtering
Memory-based learning <- memorizing the training data explicitly
Hebbian learning <- neurobiological
Competitive learning <- neurobiological
Boltzmann learning <- statistical mechanics

9
Error-Correction Learning
10
error signal = desired
response output signal
e
k
(n) = d
k
(n) y
k
(n)
e
k
(n) actuates a control
mechanism to make the
output signal y
k
(n) come
closer to the desired
response d
k
(n) in step by
step manner
A cost function c(n) = e
k
(n) is the instantaneous value of the
error energy
= a delta rule or Widrow-Hoff rule
Aw
kj
(n) = q e
k
(n) x
j
(n),
q is the learning rate parameter
The adjustment made to a synaptic weight of a neuron is
proportional to the product of the error signal and the input
signal of the synapse in question.
w
kj
(n+1) = w
kj
(n) + Aw
kj
(n)

11
Memory-Based Learning
all of the past experiences are explicitly stored in a large memory
of correctly classified input-output examples
{(x
i
,d
i
)}
N

i=1
Criterion used for defining the local neighbourhood of the test
vector x
test
.
Learning rule applied to the training examples in the local
neighborhood of x
test
.
Nearest neighbor rule: the vector x
N
e{x
1
,x
2
,...,x
N
} is the nearest
neighbor of x
test
if min
i
d(x
i
, x
test
) = d(x
N
, x
test
)

12
13
k-nearest neighbor classifier:
Identify the k classified
patterns that lie nearest to
the test vector x
test
for some
integer k.
Assign x
test
to the class that is
most frequently represented
in the k nearest neighbors to
x
test
.
Hebbian Learning:
1. If two neurons on either side of synapse (connection) are
activated simultaneously, then the strength of that synapse is
selectively increased.
2. If two neurons on either side of a synapse are activated
asynchronously, then that synapse is selectively weakened or
eliminated.
A Hebbian synapse increases its strength with positively
correlated presynaptic and postsynaptic signals, and decreases its
strength when signals are either uncorrelated or negatively
correlated.

14
15
The Hebbian learning in
matematical terms:
Aw
kj
(n)=F(y
k
(n),x
j
(n))
The simplest form:
Aw
kj
(n) = qy
k
(n)x
j
(n)
Covariance hypothesis:
Aw
kj
= q(x
j
-x)(y
j
-y)

1. Synaptic weight w
kj
is enhanced if the conditions x
j
>x and
y
k
>y are both satisfied.
2. Synaptic weight w
kj
is depressed if there is x
j
>x and y
k
<y
or y
k
>y and x
j
<x .

Competitive Learning:
16
The output neurons of a
neural network compete
among themselves to
become active.
- a set of neurons that are all
the same (excepts for
synaptic weights)
- a limit imposed on the
strength of each neuron
- a mechanism that permits
the neurons to compete -> a
winner-takes-all
The standard competitive
learning rule
Aw
kj
= q(x
j
-w
kj
) if neuron k wins
the competition
= 0 if neuron k loses the
competition

Boltzmann Learning:
The neurons constitute a recurrent structure and they operate
in a binary manner. The machine is characterized by an energy
function E.
E = -E
j
E
k
w
kj
x
k
x
j
, j=k
Machine operates by choosing a neuron at random then
flipping the state of neuron k from state x
k
to state x
k
at some
temperature T with probability
P(x
k
- x
k
) = 1/(1+exp(- AE
k
/T))

Clamped condition: the visible neurons are all clamped onto
specific states determined by the environment
Free-running condition: all the neurons (=visible and hidden) are
allowed to operate freely
The Boltzmann learning rule:
Aw
kj
= q(
+
kj
-
-
kj
), j=k,
note that both
+
kj
and
-
kj
range in value from 1 to +1.

17
18
Learning with a Teacher
(supervised learning)
Here both the inputs and the
outputs are provided.
The network then processes the
inputs and compares its resulting
outputs against the desired
outputs.
Errors are then calculated, causing
the system to adjust the weights
which control the network.
This process occurs over and over
as the weights are continually
tweaked.
The teacher has knowledge of the
environment
Learning Methods
19
Learning without a Teacher:
no labeled examples available
of the function to be learned.
Reinforcement learning:
The learning of input-output
mapping is performed
through continued
interaction with the
environment in oder to
minimize a scalar index of
performance
20
Delayed reinforcement, which means that the
system observes a temporal sequence of
stimuli.

Difficult to perform for two reasons:
-There is no teacher to provide a desired
response at each step of the learning process.

-The delay incurred in the generation of the
primary reinforcement signal implies that the
machine must solve a temporal credit
assignment problem.

Reinforcement learning is closely related to
dynamic programming.
21
Unsupervised Learning

There is no external teacher or critic to oversee the learning process.

The provision is made for a task independent measure of the quality of
representation that the network is required to learn.

Network Architectures
Single Layer Feed Forward
Networks
An input layer of source nodes
that projects onto an output
layer of neurons
Single layer refers to the output
layer of computation nodes.
Multilayer Feed Forward Net
One or more hidden layer
Adding hidden layer enables to
extract higher order statistics
Source nodes second layer Third
Layer
Output signal of output layer contains
overall response of the network.
22
Recurrent Networks
Has at least one feedback loop
Output of a neuron fedback to its input neuron
Output of a neuron fedback into its own input (Self
feedback)
Feedback loops uses unit delay elements z
-1

23

24

25
26


Neural Network types can be classified based on following attributes:

Applications
-Classification
-Clustering
-Function approximation
-Prediction
Connection Type
- Static (feedforward)
- Dynamic (feedback)
Topology
- Single layer
- Multilayer
- Recurrent
- Self-organized

Learning Methods
- Supervised
- Unsupervised
Types of Neural Networks

27

28

29

30

31

32

33

34

35

You might also like