Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

Machine Learning & Pattern Recognition

Basics of Artificial Neural Network

Dr. S. S. Chowhan

School of Computational Sciences


S. R. T. M. University, Nanded

01, November 2014

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 1 / 20
Outline of Course

1 About Brain

2 Biological Neuron

3 Benefits of ANN

4 Definition of ANN

5 Artificial / Mathematical Model of Neuron


Descripition of ANN Model
Architecture of Artificial Neural Network

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 2 / 20
Outline of Course

1 About Brain

2 Biological Neuron

3 Benefits of ANN

4 Definition of ANN

5 Artificial / Mathematical Model of Neuron


Descripition of ANN Model
Architecture of Artificial Neural Network

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 2 / 20
Outline of Course

1 About Brain

2 Biological Neuron

3 Benefits of ANN

4 Definition of ANN

5 Artificial / Mathematical Model of Neuron


Descripition of ANN Model
Architecture of Artificial Neural Network

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 2 / 20
Outline of Course

1 About Brain

2 Biological Neuron

3 Benefits of ANN

4 Definition of ANN

5 Artificial / Mathematical Model of Neuron


Descripition of ANN Model
Architecture of Artificial Neural Network

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 2 / 20
Outline of Course

1 About Brain

2 Biological Neuron

3 Benefits of ANN

4 Definition of ANN

5 Artificial / Mathematical Model of Neuron


Descripition of ANN Model
Architecture of Artificial Neural Network

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 2 / 20
About Brain

Figure 1: Human Brain

Highly complex, nonlinear and parallel computer. Capability of organizing


neurons so as perform Perceptual Pattern Recognition (recognizing a
familiar face embedded in an unfamiliar scene) in 100 to 200ms. Many
times faster than fastest digital computer in existence today

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 3 / 20
Biological Neuron

Figure 2: Biological Neuron

Signals are transmitted between neurons by electrical pulses


(action-potentials or ‘spike’ trains) traveling along the axon. These pulses
impinge on the afferent neuron at terminals called synapses. These are
found principally on a set of branching processes emerging from the cell
body (soma) known as dendrites. Each pulse occurring at a synapse
initiates the release of a small amount of chemical substance or
neurotransmitter which travels across the synaptic cleft and which is then
received at post-synaptic receptor sites on the dendrite side of the synapse.

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 4 / 20
Benefits of Neural Network

(I) Parallel Distributed Massive Structure


(II) VLSI Implementation
(III) Fault Tolerance
(IV) Uniformity of Analysis and Design
(V) Generalization
(VI) Adaptability
(VII) Non-linearity
(VIII) Input-Output Mapping

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 5 / 20
Definition of Neural Network

ANN is a massively parallel distributed processor which can store


experiential knowledge and making it available for use. It resembles the
brain in two respects:
(I) Knowledge is acquired through learning
(II) Synaptic weights are used to store the knowledge

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 6 / 20
Artificial / Mathematical Model of Neuron

Figure 3: ANN Model

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 7 / 20
Artificial / Mathematical Model of Neuron

Figure 4: Artificial Neuron Model

Fundamental Processing Unit a


Three basic elements of the neuron model

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 8 / 20
Discription of ANN Model

(a) Set of synapses or connecting links


(b) An adder for summing the input signals
(c) An activation function
(d) Threshold input lowers the net input of the neuron (Bias negative of
threshold)
(e) Adder acts as linear combiner and adds weighted inputs
(f) Activation function limits the output of a neuron
(g) Typically normalized output is [0,1] or [1,-1]

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 9 / 20
Activation Functions

Figure 5: ANN Model

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 10 / 20
net = X1 ∗ W1 + X2 ∗ W2 + ..... + Xn ∗ Wn >= t (1)
but this is equivalent algebraically equivalent to

net = X1 ∗ W1 + X2 ∗ W2 + ..... + Xn ∗ Wn + (−1) ∗ t >= 0 (2)

So, in an implementation, Consider each TLU ( Threshold Logic Unit)


having an extra input value -1 and the weight is t (threshold) a
As Shown in the below

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 11 / 20
Linear and Non-linearly separable problems

1 The two classes are therefore separated by the 0 decisionline0 which is


defined by putting the activation equal to the threshold. It turns out
that it is possible to generalize this result to TLUs with n inputs in
3-D the two classes are separated by a decision-plane. In n-D this
becomes a decision − hyperplane.
(The ‘hyper-’ is sometimes dropped even when n > 3 ).
2 The converse of this is that, any TLU is defined by some hyperplane
in its pattern space and any function which cannot be realized in this
way cannot be realized by a TLU. Because the defining equation for
the hyper plane is linear the TLU is a linear classifier.

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 12 / 20
Linear and Non-Linear Separable Problems

Figure 6: XOR Non-Linear Separable Problem

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 13 / 20
Linear and Non-Linear Separable Problems

Figure 7: Linear Separable Problem

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 14 / 20
Neural Network Architecture / Neural Network Topologies

1 FeedforwardNetwork
Singlelayer Feedforward Network
Multilayer Feedforward Network
2 FeedbackNetwork

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 15 / 20
SingleLayer Neural Network Architecture

A single layer network is a simple structure consisting of m neurons


each having n inputs
The system performs a mapping from the n-dimensional input space
to the m-dimensional output space
To train the network the same learning algorithms as for a single
neuron can be used
This type of network is widely used for linearseparable problems,
Singlelayer network are not capable of classifying
non linear separable data sets.

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 16 / 20
Multilayer Neural Network Architecture

To Solve Non-Linear Separable Problems Multilayer Neural Network


are used
This network is distinguished by the presence of one or more hidden
layers,
Whose computation nodes are hidden neurons
By adding one or more hidden layers, the network is enabled to
extract higher-order statistics and suitable for nonlilearly separable
pattern classification problems

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 17 / 20
Figure 8: Toplogy of ANN

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 18 / 20
References
Marget H. D. and S. Sridhar
Data mining : Introductory and Advanced Topics
Pearson Publishing New Delhi, 2003.
C. Apte and S. Weiss.
Data mining with decision trees and decision rules
Future Generation Computer Systems, 13, 1997.
S. K. Murthy,
Automatic Construction of Decision Trees from Data: A
Multi-Disciplinary Survey, Data Mining and Knowledge Discovery
2(4): 345-389, 1998
J. R. Quinlan.
Induction of decision trees. Machine Learning, 1:81-106, 1986.
I. H. Witten and E. Frank.
Data Mining: Practical Machine Learning Tools and Techniques, 2ed.
Morgan Kaufmann, 2005.
Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 19 / 20
References

Jack.M. Zurada
Introduction to Artificial Neural Systems.
Jaico Publishing Home, 2002.
George J. Klir, Bo Yuan
Fuzzy Sets and Fuzzy Logic Theory and Application
PHI Press 1998.
Satish Kumar
Neural Networks A CLASS ROOM APPROACH
Tata McGraw Hill Publishing 2003.
Robert J.Schalkoff
Artifical Neural Networks
Pearson Education, New Delhi, Tata McGraw Hill Publishing 1997.

Dr. S. S. Chowhan (SCS) Machine Learning & Pattern Recognition 01, November 2014 20 / 20

You might also like