PERCEPTRON

What is a Neural Network?

● A neural network can be defined as a model of reasoning based on the human

brain. The brain consists of a densely interconnected set of nerve cells, or basic
information-processing units, called neurons.
ANALOGY BETWEEN BIOLOGICAL
AND ARTIFICIAL NNS
 An Artificial Neuron
Input Signal

x1 w1
Output Signal
Neuron

x2 w2

Σ Y Y

x3 w3

θ
xn wn
 Single Layer Neural Net
w11
w21

y
x

1
1
wi1
wn1

w12

Output signal
w22
Input Signal

y
x

wi2

2
2

wn2

· ·
· w13
·
· w23 ·
wi3

yj
i
x

wnj

w1m
·
·
·

∙
∙
∙
w2m
wim
wnm ym
xn

Input Output
layer layer
 Multilayer Neural Net
x1
V11 Z1
V1m w11

Output Signals
Input Signals

x2 V21 y1 w12
V2m w1n
Z2
V31 wm1
x3 ym wm2
V3m

wmn Zn
vl1
Vlm
xl
 CHARACTERISTICS OF NN

▶ The NN exhibit mapping capabilities.

▶ The NNs learn by examples.
▶ The NNs process the capability to generalize
▶ The NN robust systems and are fault tolerant
▶ The NNs can process information in parallel, at
high speed, and in a distributed manner.
 Activation Function

•The activation function is used to calculate the output

response of a neuron.
• The sum of the weighted input signal is applied with
an activation to obtain there response.
•For neurons in same layer, same activation functions
are used.
Activation Function of Neuron
 Step Activation Function

f(x) f(x)
1 1

θ=0 x 0 x

-5 -4 -3 -2 -1 0 1 2 3 4 5 -5 -4 -3 -2 -1 0 1 2 3 4 5
θ=1
 Sign Activation Function

f(x)

+1

θ=0 x
-5 -4 -3 -2 -1 0 1 2 3 4 5

-1-1
-1
 Sigmoid Activation Function
Binary
Sigmoid Bipolar Sigmoid
f(x)

f(x)

x
0 +1

When x=0,
-1
 Activation Function
Find the output of the following net when (i) Step function (θ =0), (ii) Step
function (θ =0.5), (iii) Step function (θ =1.0), (iv) Sign function (θ =-0.5), (v) Sign
function (θ =1.0), (vi) Binary sigmoid function (vii) Bipolar sigmoid function are
used.
x1
0.2
1.0
0.3 x2 -0.3
0.2
-0.1
0.7 x3 0.2 ∑ y

0.4

x4
0.5
0.5

x5
0.1
 Activation Function(Solution)
Find the output of the following net when (i) Step function (θ =0), are used.

x1
0.2
1.0
0.3
x2 -0.3
0.2
-0.1
0.7 x3 0.2 ∑ Z

Z=Step(0.2*(-o.3)+0.3*(-0.1)+0.7*0.2+1.0*0.2)
Z= Step( -0.09+0.14+0.2)
Z=Step(0.25)
Z=1
TYPES OF ARTIFICIAL NEURAL NETWORK

▶ Perceptron
▶ BAM(Bidirectional Associative Memory)
▶ FAM(Fuzzy Associative Memory )
▶ Hebb Net
▶ Hopfield Network
▶ SOFM(Self Organizing Neural Net)
 APPLICATIONS OF ANN

▶ Medical Diagnosis
▶ Assisting doctors with their diagnosis by analyzing the
reported symptoms.
▶ Input data: patient's personal information, patterns of
symptoms, heart rate, blood pressure, temperature,
laboratory results, treatment procedure etc.

Patient's Length of Stay Forecasts

forecast which patients remain for a specified
number of days.
 APPLICATIONS OF ANN

Prediction
▶ -use some variables or fields in the
database to predict unknown or future
values of other variables of interest.
Knowledge Discovery
- find new relationships and
non-obvious trends in the data.
Quality Control
- predict the quality of plastics, paper,
and other raw materials
 APPLICATIONS OF ANN

Electronic noses
• Electronic noses have several potential
applications in telemedicine.
• Telemedicine is the practice of medicine over
long distances via a communication link.
• The electronic nose would identify smells in the
remote surgical environment.
• These identified smells would then be
electronically transmitted to another site where an
door generation system would recreate them.
• Because the sense of smell can be an important
sense to the surgeon, telesmell would enhance
telepresent surgery.
LEARNING OF NEURAL
NETWORK

▶ Learning is process by which a neural network

adapts itself to a stimulus by properly making
parameter adjustment and producing a desired
response
▶ Types
▶ Supervised learning
▶ Unsupervised learning
SUPERVISED LEARNING

❖Child learns from a teacher

❖Each input vector requires a corresponding target vector.

❖ Training pair=[input vector, target vector]

Neural
X Network Y
W
(Input) (Actual output)
Error
Error
(D-Y) Signal
signals Generator (Desired Output)
UNSUPERVISED LEARNING

Unsupervised training requires no teacher; input patterns are applied, and the
network self-organizes by adjusting its weights according to a well-defined
algorithm. Since no desired output is specified during the training process, the
results are unpredictable in terms of firing patterns of specific neurons. In this
method of training, the input vectors of similar types are grouped without the
use of training data to specify how a typical member of each group looks or to
which group a member belongs. During training the neural network receives
input patterns and organizes these patterns into categories. When new input
pattern is applied, the neural network provides an output response indicating
the class to which the input pattern belongs. If a class cannot be found for the
input pattern, a new class is generated.
 Single Layer Perceptron
Perceptron is the simplest form of a neural network.
It consists of a single neuron with adjustable synaptic
weights and a hard limiter activation function.

Linear Hard
x1 Combiner Limiter
W1
Output
Input Σ Y
s
W2
θ

x2
 ROSENBLATT’S PERCEPTRON

Step 1: Initialization
Set initial weights w1, w2, ..wn and threshold θ to random numbers [-0.5,0.5].
Step 2: Activation
Activate the perceptron by applying inputs x1(p), x2(p), x3(p), ……xn(p) and
desired output Yd(p). Calculate the actual output at iteration p=1:

Where n is the number of the perceptron inputs, and step is a step activation
function.
Step 3: Weight training
Update the weights of the perceptron

Step 4: Iteration
Increase iteration p by one, go back to Step 2 and repeat the process until