neural network · 2020. 4. 22. · structure of biological neuron biological neuron cell axonsare...

19
Neural Network Ashok N Shinde [email protected] International Institute of Information Technology Hinjawadi Pune December 15, 2017 Ashok N Shinde Neural Network 1/19

Upload: others

Post on 19-Sep-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Neural Network

Ashok N Shindeashokshinde0349gmailcom

International Institute of Information TechnologyHinjawadi Pune

December 15 2017

Ashok N Shinde Neural Network 119

Introduction

Introduction to Neural Network

What is a Neural Network

Benefits of Neural Networks

The Human Nervous System

The Human Brain

Structure of Biological Neuron

Models of A Neuron

Neural Networks and Graphs

Feedback

Network Architectures

Knowledge Representation

Learning Processes

Learning Tasks

Ashok N Shinde Neural Network 219

What is a Neural Network

What is a Neural Network

A neural network is

A massively parallel distributed processor made up of simpleprocessing units called neuronsHas tendency for storing knowledge and making it available foruseHas ability to change its structure and modify according toapplication requirement

It resembles the brain in two respects

Knowledge is acquired by the network from its environmentthrough the learning processInter neuron connection strengths known as synaptic weightsare used to store the acquired knowledge

Ashok N Shinde Neural Network 319

Benefits of Neural Networks

Nonlinearity (NN could be linear or nonlinear)

A highly important property particularly if the underlyingphysical mechanism responsible for generation of the inputsignal (eg speech signal) is inherently nonlinear

Input-Output Mapping

Knowledge is acquired by the network from its environmentthrough the learning processIt finds the optimal mappingbetween an input signal and the output results throughlearning mechanism that adjust the weights that minimize thedifference between the actual response and the desired one(non parametric statistical inference)

Adaptivity

A neural network could be designed to change its weights inreal time which enables the system to operate in a nonstationary environment

Ashok N Shinde Neural Network 419

Benefits of Neural Networks

Evidential Response

In the context of pattern classification a network can bedesigned to provide information not only about whichparticular pattern to select but also about the confidence inthe decision made

Contextual Information

Every neuron in the network is potentially affected by theglobal activity of all other neurons Consequently contextualinformation is dealt with naturally by a neural network

Fault tolerance

A neural network is inherently fault tolerant or capable ofrobust computation in the sense that its performancedegrades gracefully under adverse operating conditions

Ashok N Shinde Neural Network 519

Benefits of Neural Networks

VLSI Implementation

The massively parallel nature of a neural network makes itpotentially fast for a certain computation task which alsomakes it well suited for implementation using VLSI technology

Uniformity of Analysis and DesignNeural networks enjoy universality as information processors

Neurons in one form or another represent an ingredientcommon to all neural networks which makes it possible toshare theories and learning algorithms in different applicationsModular networks can be built through a seamless integrationof modules

Neurobiology Analogy

The design of a neural network is motivated by analogy to thebrain (the fastest and powerful fault tolerant parallelprocessor)

Ashok N Shinde Neural Network 619

Application of Neural Networks

Neural networks are tractable and easy to implementSpecially in hardware This made it attractive to be used in awide range of applications

Pattern ClassificationsMedical ApplicationsForecastingAdaptive FilteringAdaptive Control

Ashok N Shinde Neural Network 719

The Human Nervous System

The human nervous system may be viewed as a three-stagesystem

The brain represented by the neural net is central to thesystem It continually receive information perceives it andmakes appropriate decisionThe receptors convert stimuli from the human body or theexternal environment into electrical impulses that conveyinformation to the brainThe effectors convert electrical impulses generated by the braininto distinct responses as system output

Ashok N Shinde Neural Network 819

The Human Brain

There are approximately 10 billion neurons in the humancortex compared with 10 of thousands of processors in themost powerful parallel computers

Each biological neuron is connected to several thousands ofother neurons

The typical operating speeds of biological neurons is measuredin milliseconds (10minus3s) while a silicon chip can operate innanoseconds (10minus9s) (Lack of processing units in computerscan be compensated by speed)

The human brain is extremely energy efficient usingapproximately 10minus16 joules per operation per second whereasthe best computers today use around 10minus6 joules peroperation per second

Ashok N Shinde Neural Network 919

Structure of Biological Neuron

Biological Neuron cell

The neurons cell body (soma) processes the incomingactivations and converts them into output activations

Dendrites are fibers which emanate from the cell body andprovide the receptive zones that receive activation from otherneurons

Ashok N Shinde Neural Network 1019

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 2: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Introduction

Introduction to Neural Network

What is a Neural Network

Benefits of Neural Networks

The Human Nervous System

The Human Brain

Structure of Biological Neuron

Models of A Neuron

Neural Networks and Graphs

Feedback

Network Architectures

Knowledge Representation

Learning Processes

Learning Tasks

Ashok N Shinde Neural Network 219

What is a Neural Network

What is a Neural Network

A neural network is

A massively parallel distributed processor made up of simpleprocessing units called neuronsHas tendency for storing knowledge and making it available foruseHas ability to change its structure and modify according toapplication requirement

It resembles the brain in two respects

Knowledge is acquired by the network from its environmentthrough the learning processInter neuron connection strengths known as synaptic weightsare used to store the acquired knowledge

Ashok N Shinde Neural Network 319

Benefits of Neural Networks

Nonlinearity (NN could be linear or nonlinear)

A highly important property particularly if the underlyingphysical mechanism responsible for generation of the inputsignal (eg speech signal) is inherently nonlinear

Input-Output Mapping

Knowledge is acquired by the network from its environmentthrough the learning processIt finds the optimal mappingbetween an input signal and the output results throughlearning mechanism that adjust the weights that minimize thedifference between the actual response and the desired one(non parametric statistical inference)

Adaptivity

A neural network could be designed to change its weights inreal time which enables the system to operate in a nonstationary environment

Ashok N Shinde Neural Network 419

Benefits of Neural Networks

Evidential Response

In the context of pattern classification a network can bedesigned to provide information not only about whichparticular pattern to select but also about the confidence inthe decision made

Contextual Information

Every neuron in the network is potentially affected by theglobal activity of all other neurons Consequently contextualinformation is dealt with naturally by a neural network

Fault tolerance

A neural network is inherently fault tolerant or capable ofrobust computation in the sense that its performancedegrades gracefully under adverse operating conditions

Ashok N Shinde Neural Network 519

Benefits of Neural Networks

VLSI Implementation

The massively parallel nature of a neural network makes itpotentially fast for a certain computation task which alsomakes it well suited for implementation using VLSI technology

Uniformity of Analysis and DesignNeural networks enjoy universality as information processors

Neurons in one form or another represent an ingredientcommon to all neural networks which makes it possible toshare theories and learning algorithms in different applicationsModular networks can be built through a seamless integrationof modules

Neurobiology Analogy

The design of a neural network is motivated by analogy to thebrain (the fastest and powerful fault tolerant parallelprocessor)

Ashok N Shinde Neural Network 619

Application of Neural Networks

Neural networks are tractable and easy to implementSpecially in hardware This made it attractive to be used in awide range of applications

Pattern ClassificationsMedical ApplicationsForecastingAdaptive FilteringAdaptive Control

Ashok N Shinde Neural Network 719

The Human Nervous System

The human nervous system may be viewed as a three-stagesystem

The brain represented by the neural net is central to thesystem It continually receive information perceives it andmakes appropriate decisionThe receptors convert stimuli from the human body or theexternal environment into electrical impulses that conveyinformation to the brainThe effectors convert electrical impulses generated by the braininto distinct responses as system output

Ashok N Shinde Neural Network 819

The Human Brain

There are approximately 10 billion neurons in the humancortex compared with 10 of thousands of processors in themost powerful parallel computers

Each biological neuron is connected to several thousands ofother neurons

The typical operating speeds of biological neurons is measuredin milliseconds (10minus3s) while a silicon chip can operate innanoseconds (10minus9s) (Lack of processing units in computerscan be compensated by speed)

The human brain is extremely energy efficient usingapproximately 10minus16 joules per operation per second whereasthe best computers today use around 10minus6 joules peroperation per second

Ashok N Shinde Neural Network 919

Structure of Biological Neuron

Biological Neuron cell

The neurons cell body (soma) processes the incomingactivations and converts them into output activations

Dendrites are fibers which emanate from the cell body andprovide the receptive zones that receive activation from otherneurons

Ashok N Shinde Neural Network 1019

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 3: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

What is a Neural Network

What is a Neural Network

A neural network is

A massively parallel distributed processor made up of simpleprocessing units called neuronsHas tendency for storing knowledge and making it available foruseHas ability to change its structure and modify according toapplication requirement

It resembles the brain in two respects

Knowledge is acquired by the network from its environmentthrough the learning processInter neuron connection strengths known as synaptic weightsare used to store the acquired knowledge

Ashok N Shinde Neural Network 319

Benefits of Neural Networks

Nonlinearity (NN could be linear or nonlinear)

A highly important property particularly if the underlyingphysical mechanism responsible for generation of the inputsignal (eg speech signal) is inherently nonlinear

Input-Output Mapping

Knowledge is acquired by the network from its environmentthrough the learning processIt finds the optimal mappingbetween an input signal and the output results throughlearning mechanism that adjust the weights that minimize thedifference between the actual response and the desired one(non parametric statistical inference)

Adaptivity

A neural network could be designed to change its weights inreal time which enables the system to operate in a nonstationary environment

Ashok N Shinde Neural Network 419

Benefits of Neural Networks

Evidential Response

In the context of pattern classification a network can bedesigned to provide information not only about whichparticular pattern to select but also about the confidence inthe decision made

Contextual Information

Every neuron in the network is potentially affected by theglobal activity of all other neurons Consequently contextualinformation is dealt with naturally by a neural network

Fault tolerance

A neural network is inherently fault tolerant or capable ofrobust computation in the sense that its performancedegrades gracefully under adverse operating conditions

Ashok N Shinde Neural Network 519

Benefits of Neural Networks

VLSI Implementation

The massively parallel nature of a neural network makes itpotentially fast for a certain computation task which alsomakes it well suited for implementation using VLSI technology

Uniformity of Analysis and DesignNeural networks enjoy universality as information processors

Neurons in one form or another represent an ingredientcommon to all neural networks which makes it possible toshare theories and learning algorithms in different applicationsModular networks can be built through a seamless integrationof modules

Neurobiology Analogy

The design of a neural network is motivated by analogy to thebrain (the fastest and powerful fault tolerant parallelprocessor)

Ashok N Shinde Neural Network 619

Application of Neural Networks

Neural networks are tractable and easy to implementSpecially in hardware This made it attractive to be used in awide range of applications

Pattern ClassificationsMedical ApplicationsForecastingAdaptive FilteringAdaptive Control

Ashok N Shinde Neural Network 719

The Human Nervous System

The human nervous system may be viewed as a three-stagesystem

The brain represented by the neural net is central to thesystem It continually receive information perceives it andmakes appropriate decisionThe receptors convert stimuli from the human body or theexternal environment into electrical impulses that conveyinformation to the brainThe effectors convert electrical impulses generated by the braininto distinct responses as system output

Ashok N Shinde Neural Network 819

The Human Brain

There are approximately 10 billion neurons in the humancortex compared with 10 of thousands of processors in themost powerful parallel computers

Each biological neuron is connected to several thousands ofother neurons

The typical operating speeds of biological neurons is measuredin milliseconds (10minus3s) while a silicon chip can operate innanoseconds (10minus9s) (Lack of processing units in computerscan be compensated by speed)

The human brain is extremely energy efficient usingapproximately 10minus16 joules per operation per second whereasthe best computers today use around 10minus6 joules peroperation per second

Ashok N Shinde Neural Network 919

Structure of Biological Neuron

Biological Neuron cell

The neurons cell body (soma) processes the incomingactivations and converts them into output activations

Dendrites are fibers which emanate from the cell body andprovide the receptive zones that receive activation from otherneurons

Ashok N Shinde Neural Network 1019

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 4: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Benefits of Neural Networks

Nonlinearity (NN could be linear or nonlinear)

A highly important property particularly if the underlyingphysical mechanism responsible for generation of the inputsignal (eg speech signal) is inherently nonlinear

Input-Output Mapping

Knowledge is acquired by the network from its environmentthrough the learning processIt finds the optimal mappingbetween an input signal and the output results throughlearning mechanism that adjust the weights that minimize thedifference between the actual response and the desired one(non parametric statistical inference)

Adaptivity

A neural network could be designed to change its weights inreal time which enables the system to operate in a nonstationary environment

Ashok N Shinde Neural Network 419

Benefits of Neural Networks

Evidential Response

In the context of pattern classification a network can bedesigned to provide information not only about whichparticular pattern to select but also about the confidence inthe decision made

Contextual Information

Every neuron in the network is potentially affected by theglobal activity of all other neurons Consequently contextualinformation is dealt with naturally by a neural network

Fault tolerance

A neural network is inherently fault tolerant or capable ofrobust computation in the sense that its performancedegrades gracefully under adverse operating conditions

Ashok N Shinde Neural Network 519

Benefits of Neural Networks

VLSI Implementation

The massively parallel nature of a neural network makes itpotentially fast for a certain computation task which alsomakes it well suited for implementation using VLSI technology

Uniformity of Analysis and DesignNeural networks enjoy universality as information processors

Neurons in one form or another represent an ingredientcommon to all neural networks which makes it possible toshare theories and learning algorithms in different applicationsModular networks can be built through a seamless integrationof modules

Neurobiology Analogy

The design of a neural network is motivated by analogy to thebrain (the fastest and powerful fault tolerant parallelprocessor)

Ashok N Shinde Neural Network 619

Application of Neural Networks

Neural networks are tractable and easy to implementSpecially in hardware This made it attractive to be used in awide range of applications

Pattern ClassificationsMedical ApplicationsForecastingAdaptive FilteringAdaptive Control

Ashok N Shinde Neural Network 719

The Human Nervous System

The human nervous system may be viewed as a three-stagesystem

The brain represented by the neural net is central to thesystem It continually receive information perceives it andmakes appropriate decisionThe receptors convert stimuli from the human body or theexternal environment into electrical impulses that conveyinformation to the brainThe effectors convert electrical impulses generated by the braininto distinct responses as system output

Ashok N Shinde Neural Network 819

The Human Brain

There are approximately 10 billion neurons in the humancortex compared with 10 of thousands of processors in themost powerful parallel computers

Each biological neuron is connected to several thousands ofother neurons

The typical operating speeds of biological neurons is measuredin milliseconds (10minus3s) while a silicon chip can operate innanoseconds (10minus9s) (Lack of processing units in computerscan be compensated by speed)

The human brain is extremely energy efficient usingapproximately 10minus16 joules per operation per second whereasthe best computers today use around 10minus6 joules peroperation per second

Ashok N Shinde Neural Network 919

Structure of Biological Neuron

Biological Neuron cell

The neurons cell body (soma) processes the incomingactivations and converts them into output activations

Dendrites are fibers which emanate from the cell body andprovide the receptive zones that receive activation from otherneurons

Ashok N Shinde Neural Network 1019

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 5: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Benefits of Neural Networks

Evidential Response

In the context of pattern classification a network can bedesigned to provide information not only about whichparticular pattern to select but also about the confidence inthe decision made

Contextual Information

Every neuron in the network is potentially affected by theglobal activity of all other neurons Consequently contextualinformation is dealt with naturally by a neural network

Fault tolerance

A neural network is inherently fault tolerant or capable ofrobust computation in the sense that its performancedegrades gracefully under adverse operating conditions

Ashok N Shinde Neural Network 519

Benefits of Neural Networks

VLSI Implementation

The massively parallel nature of a neural network makes itpotentially fast for a certain computation task which alsomakes it well suited for implementation using VLSI technology

Uniformity of Analysis and DesignNeural networks enjoy universality as information processors

Neurons in one form or another represent an ingredientcommon to all neural networks which makes it possible toshare theories and learning algorithms in different applicationsModular networks can be built through a seamless integrationof modules

Neurobiology Analogy

The design of a neural network is motivated by analogy to thebrain (the fastest and powerful fault tolerant parallelprocessor)

Ashok N Shinde Neural Network 619

Application of Neural Networks

Neural networks are tractable and easy to implementSpecially in hardware This made it attractive to be used in awide range of applications

Pattern ClassificationsMedical ApplicationsForecastingAdaptive FilteringAdaptive Control

Ashok N Shinde Neural Network 719

The Human Nervous System

The human nervous system may be viewed as a three-stagesystem

The brain represented by the neural net is central to thesystem It continually receive information perceives it andmakes appropriate decisionThe receptors convert stimuli from the human body or theexternal environment into electrical impulses that conveyinformation to the brainThe effectors convert electrical impulses generated by the braininto distinct responses as system output

Ashok N Shinde Neural Network 819

The Human Brain

There are approximately 10 billion neurons in the humancortex compared with 10 of thousands of processors in themost powerful parallel computers

Each biological neuron is connected to several thousands ofother neurons

The typical operating speeds of biological neurons is measuredin milliseconds (10minus3s) while a silicon chip can operate innanoseconds (10minus9s) (Lack of processing units in computerscan be compensated by speed)

The human brain is extremely energy efficient usingapproximately 10minus16 joules per operation per second whereasthe best computers today use around 10minus6 joules peroperation per second

Ashok N Shinde Neural Network 919

Structure of Biological Neuron

Biological Neuron cell

The neurons cell body (soma) processes the incomingactivations and converts them into output activations

Dendrites are fibers which emanate from the cell body andprovide the receptive zones that receive activation from otherneurons

Ashok N Shinde Neural Network 1019

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 6: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Benefits of Neural Networks

VLSI Implementation

The massively parallel nature of a neural network makes itpotentially fast for a certain computation task which alsomakes it well suited for implementation using VLSI technology

Uniformity of Analysis and DesignNeural networks enjoy universality as information processors

Neurons in one form or another represent an ingredientcommon to all neural networks which makes it possible toshare theories and learning algorithms in different applicationsModular networks can be built through a seamless integrationof modules

Neurobiology Analogy

The design of a neural network is motivated by analogy to thebrain (the fastest and powerful fault tolerant parallelprocessor)

Ashok N Shinde Neural Network 619

Application of Neural Networks

Neural networks are tractable and easy to implementSpecially in hardware This made it attractive to be used in awide range of applications

Pattern ClassificationsMedical ApplicationsForecastingAdaptive FilteringAdaptive Control

Ashok N Shinde Neural Network 719

The Human Nervous System

The human nervous system may be viewed as a three-stagesystem

The brain represented by the neural net is central to thesystem It continually receive information perceives it andmakes appropriate decisionThe receptors convert stimuli from the human body or theexternal environment into electrical impulses that conveyinformation to the brainThe effectors convert electrical impulses generated by the braininto distinct responses as system output

Ashok N Shinde Neural Network 819

The Human Brain

There are approximately 10 billion neurons in the humancortex compared with 10 of thousands of processors in themost powerful parallel computers

Each biological neuron is connected to several thousands ofother neurons

The typical operating speeds of biological neurons is measuredin milliseconds (10minus3s) while a silicon chip can operate innanoseconds (10minus9s) (Lack of processing units in computerscan be compensated by speed)

The human brain is extremely energy efficient usingapproximately 10minus16 joules per operation per second whereasthe best computers today use around 10minus6 joules peroperation per second

Ashok N Shinde Neural Network 919

Structure of Biological Neuron

Biological Neuron cell

The neurons cell body (soma) processes the incomingactivations and converts them into output activations

Dendrites are fibers which emanate from the cell body andprovide the receptive zones that receive activation from otherneurons

Ashok N Shinde Neural Network 1019

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 7: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Application of Neural Networks

Neural networks are tractable and easy to implementSpecially in hardware This made it attractive to be used in awide range of applications

Pattern ClassificationsMedical ApplicationsForecastingAdaptive FilteringAdaptive Control

Ashok N Shinde Neural Network 719

The Human Nervous System

The human nervous system may be viewed as a three-stagesystem

The brain represented by the neural net is central to thesystem It continually receive information perceives it andmakes appropriate decisionThe receptors convert stimuli from the human body or theexternal environment into electrical impulses that conveyinformation to the brainThe effectors convert electrical impulses generated by the braininto distinct responses as system output

Ashok N Shinde Neural Network 819

The Human Brain

There are approximately 10 billion neurons in the humancortex compared with 10 of thousands of processors in themost powerful parallel computers

Each biological neuron is connected to several thousands ofother neurons

The typical operating speeds of biological neurons is measuredin milliseconds (10minus3s) while a silicon chip can operate innanoseconds (10minus9s) (Lack of processing units in computerscan be compensated by speed)

The human brain is extremely energy efficient usingapproximately 10minus16 joules per operation per second whereasthe best computers today use around 10minus6 joules peroperation per second

Ashok N Shinde Neural Network 919

Structure of Biological Neuron

Biological Neuron cell

The neurons cell body (soma) processes the incomingactivations and converts them into output activations

Dendrites are fibers which emanate from the cell body andprovide the receptive zones that receive activation from otherneurons

Ashok N Shinde Neural Network 1019

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 8: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

The Human Nervous System

The human nervous system may be viewed as a three-stagesystem

The brain represented by the neural net is central to thesystem It continually receive information perceives it andmakes appropriate decisionThe receptors convert stimuli from the human body or theexternal environment into electrical impulses that conveyinformation to the brainThe effectors convert electrical impulses generated by the braininto distinct responses as system output

Ashok N Shinde Neural Network 819

The Human Brain

There are approximately 10 billion neurons in the humancortex compared with 10 of thousands of processors in themost powerful parallel computers

Each biological neuron is connected to several thousands ofother neurons

The typical operating speeds of biological neurons is measuredin milliseconds (10minus3s) while a silicon chip can operate innanoseconds (10minus9s) (Lack of processing units in computerscan be compensated by speed)

The human brain is extremely energy efficient usingapproximately 10minus16 joules per operation per second whereasthe best computers today use around 10minus6 joules peroperation per second

Ashok N Shinde Neural Network 919

Structure of Biological Neuron

Biological Neuron cell

The neurons cell body (soma) processes the incomingactivations and converts them into output activations

Dendrites are fibers which emanate from the cell body andprovide the receptive zones that receive activation from otherneurons

Ashok N Shinde Neural Network 1019

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 9: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

The Human Brain

There are approximately 10 billion neurons in the humancortex compared with 10 of thousands of processors in themost powerful parallel computers

Each biological neuron is connected to several thousands ofother neurons

The typical operating speeds of biological neurons is measuredin milliseconds (10minus3s) while a silicon chip can operate innanoseconds (10minus9s) (Lack of processing units in computerscan be compensated by speed)

The human brain is extremely energy efficient usingapproximately 10minus16 joules per operation per second whereasthe best computers today use around 10minus6 joules peroperation per second

Ashok N Shinde Neural Network 919

Structure of Biological Neuron

Biological Neuron cell

The neurons cell body (soma) processes the incomingactivations and converts them into output activations

Dendrites are fibers which emanate from the cell body andprovide the receptive zones that receive activation from otherneurons

Ashok N Shinde Neural Network 1019

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 10: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Structure of Biological Neuron

Biological Neuron cell

The neurons cell body (soma) processes the incomingactivations and converts them into output activations

Dendrites are fibers which emanate from the cell body andprovide the receptive zones that receive activation from otherneurons

Ashok N Shinde Neural Network 1019

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 11: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Structure of Biological Neuron

Biological Neuron cell

Axons are fibers acting as transmission lines that sendactivation to other neurons

The junctions that allow signal transmission between theaxons and dendrites are called synapses The process oftransmission is by diffusion of chemicals calledneurotransmitters across the synaptic connection

Ashok N Shinde Neural Network 1119

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 12: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Brief History of ANN I

1943 McCulloch and Pitts proposed the McCulloch-Pitts neuronmodel

1949 Hebb published his book The Organization of Behavior in

which the Hebbian learning rule was proposed

1958 Rosenblatt introduced the simple single layer networks nowcalled Perceptrons

1969 Minsky and Paperts book Perceptrons demonstrated the

limitation of single layer perceptrons and almost the whole field

went into hibernation

1982 Hopfield published a series of papers on Hopfield networks

1982 Kohonen developed the Self-Organising Maps that now bear

his name

1986 The Back-Propagation learning algorithm for Multi-LayerPerceptrons was rediscovered and the whole field took off again

Ashok N Shinde Neural Network 1219

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 13: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Brief History of ANN II

1990 The sub-field of Radial Basis Function Networks was

developed

2000 The power of Ensembles of Neural Networks and Support

Vector Machines becomes apparent

Ashok N Shinde Neural Network 1319

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 14: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Models of a Neuron

The artificial neuron is made up of three basic elements

A set of synapses or connecting links each of which ischaracterized by a weight or strength of its own

An adder for summing the input signals weighted by therespective synaptic weight

An activation function for limiting the amplitude of theoutput of a neuron

Ashok N Shinde Neural Network 1419

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 15: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Models of a Neuron

In mathematical terms

The output of the summing function is the linear combineroutput uk =

summj=1wkjxj

and the final output signal of the neuron yk = ϕ(uk + bk)ϕ() is the activation function

Ashok N Shinde Neural Network 1519

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 16: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Effect of a Bias

In mathematical terms

The use of a bias bk has the effect of applying affinetransformation to the output uk vk = uk + bkThat is the bias value changes the relation between theinduced local field or the activation potential vk and thelinear combiner uk as shown in Fig

Ashok N Shinde Neural Network 1619

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 17: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Models of a Neuron

Models of a Neuron

Then we may write vk =summ

j=0wkjxj

and yk = ϕ(vk)

where x0 = +1 and w0k = bk

Ashok N Shinde Neural Network 1719

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 18: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

References I

1 S Haykin Neural Networks and Learning Machines3edPrentice Hall (Pearson) 2009

2 L Fausett Fundamental of Neural Networks ArchitecturesAlgorithms and Applications Prentice Hall 1995

3 httpssimplewikipediaorgwikiFileNeuronsvg

Ashok N Shinde Neural Network 1819

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919

Page 19: Neural Network · 2020. 4. 22. · Structure of Biological Neuron Biological Neuron cell Axonsare bers acting astransmission linesthat send activation to other neurons. The junctions

Author Information

For further information please contact

Prof Ashok N ShindeDepartment of Electronics amp Telecommunication EngineeringHope FoundationrsquosInternational Institute of Information Technology (I2IT )Hinjawadi Pune 411 057Phone - +91 20 22933441wwwisquareiteduin|ashoksisquareiteduin

Ashok N Shinde Neural Network 1919