Introduction To Artificial Neural Networks

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 19

MODULE:1

Introduction to Artificial Neural Networks


TOPICS
 Artificial Neural Networks
 Biological Neural Networks

 Characteristics of artificial neural networks

 Types of activation functions

 Models of neuron Topology


WHAT ARE NEURAL NETWORKS?
 A neural network is a massively
parallel, distributed processor made
up of simple processing units
(artificial neurons).

It resembles the brain in two aspects:


 Knowledge is acquired by the
network from its environment
through a learning process
 Synaptic connection strengths among
neurons are used to store the acquired
knowledge.
TWO TYPES OF NERVES OR NEURON
 Efferent, or motor nerve fibres carry impulses away
from the central nervous system

 Afferent, or sensory fibres carry impulses toward the


central nervous system
The Biological Neuron

 The human brain is made of about 100 billions of such


neurons.
 Neurons receive electric signals at the dendrites and
send them to the axon
NEURAL COMMUNICATION
 The axon of the neuron is connected to the dendrites of
many other neurons
NEUROTRANSMITTER

 Acetylcholine a chemical message released by nerve


cells to send signals to other cells, such as neurons,
muscle cells and gland cells.
TERMINOLOGY

Biological Artificial
Cell body Neuron

Dendrites Weights or interconnections

Soma Net input

Axon Output
HISTORICAL DEVELOPMENT
 1943 McCulloch and Pitts proposed the McCulloch-Pitts neuron model
 1949 Hebb , Hebbian learning rule was introduced
 1958 Rosenblatt introduced the simple single layer networks called Perceptrons
 1969 Minsky and Papert’s book Perceptrons demonstrated the limitation of single
layer perceptrons
 1980 Grossberg introduced Adaptive Resonance Theory (ART)
 1982 Hopfield published a series of papers on Hopfield networks
 1982 Kohonen developed the Self-Organizing Feature Maps
 1986 Back-propagation learning algorithm for multi-layer perceptrons was
rediscovered,
 1990s ART-variant networks were developed
 1990s Radial Basis Functions were developed
 2000s Support Vector Machines were developed
CHARACTERISTICS OF BIOLOGICAL NEURAL NETWORKS

1) Massive connectivity
2) Nonlinear, Parallel, Robust and Fault Tolerant
3) Capability to adapt to surroundings
4) Ability to learn and generalize from known examples
5) Collective behavior is different from individual behavior

Artificial Neural Networks mimics some of the


properties of the biological neural networks
SOME PROPERTIES OF ARTIFICIAL NEURAL NETWORKS

 Assembly of simple processors


 Information stored in connections – No Memory
 Massively Parallel
 Massive connectivity
 Fault Tolerant
 Learning and Generalization Ability
 Robust
 Individual dynamics different from group dynamics
All these properties may not be present in a particular
network
COMPARISON BETWEEN BRAIN VERSES
COMPUTER
SOME LEARNING ALGORITHMS WE
WILL LEARN ARE
 Supervised:
 Adaline, Madaline
 Perceptron

 Back Propagation

 multilayer perceptrons

 Radial Basis Function Networks

 Unsupervised
 Competitive Learning
 Kohenen self organizing map

 Learning vector quantization

 Hebbian learning
LEARNING

 Supervised : In Presence of a teacher

 Unsupervised or Self-Organized : No teacher

 Reinforcement: Trial and error, no teacher, but can asses


the situations – reinforcement signals.
Model of an Artificial Neuron
ACTIVATION FUNCTIONS
ACTIVATION FUNCTIONS
NETWORK TOPOLOGIES
 3 fundamental architectures:

 Single layer network


 Multi layer network
 Recurrent network
PROBLEM

1
-1
-1
1

You might also like