Upload
mrajkumarpatel
View
224
Download
0
Embed Size (px)
Citation preview
8/10/2019 Introduction ANN
1/19
ECE407 NEURAL NETWORKS & FUZZYCONTROL
8/10/2019 Introduction ANN
2/19
8/10/2019 Introduction ANN
3/19
Reference Books
8/10/2019 Introduction ANN
4/19
Mode of Evaluation
Component Weightage
CAT I 15
CAT II 15
Quiz I 5Quiz II 5
Assignment 10
8/10/2019 Introduction ANN
5/19
Guidelines for Assignment
Option 1:Simulation of any algorithm discussed in theclass. It has to be application oriented not a merenumerical problem.
Option 2 :Refer a journal paper and explain the algorithmand how it is applied . Also Solve three numericalproblems other than what we have solved in theclass and attach a copy of the journal paper
8/10/2019 Introduction ANN
6/19
Guidelines for Assignments
Priory inform the topic to avoid repetition. Not more than two per team.
It has to be submitted along with a report It has to submitted in time. All the members should come for submission Short Viva-Voce will be taken individually
during the submission.
8/10/2019 Introduction ANN
7/19
8/10/2019 Introduction ANN
8/19
Introduction
Why ANN Some tasks can be done easily (effortlessly) by humans but
are hard by conventional paradigms on Von Neumannmachine with algorithmic approach
Pattern recognition (old friends, hand-written characters) Content addressable recall Approximate, common sense reasoning (driving, playing
piano, baseball player) These tasks are often ill-defined, experience based, hard to
apply logic
8/10/2019 Introduction ANN
9/19
Introduction
Von Neumann machine -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- One or a few high speed (ns)
processors with considerablecomputing power
One or a few shared highspeed buses forcommunication
Sequential memory access byaddress
Problem-solving knowledge isseparated from the computingcomponent
Hard to be adaptive
Human Brain-------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Large # (10 11 ) of low speed
processors (ms) with limitedcomputing power
Large # (10 15) of low speedconnections
Content addressable recall(CAM)
Problem-solving knowledgeresides in the connectivity ofneurons
Adaptation by changing theconnectivity
8/10/2019 Introduction ANN
10/19
Biological neural activity
Each neuron has a body , an axon , and many dendrites Can be in one of the two states: firing and rest. Neuron fires if the total incoming stimulus exceeds the threshold
Synapse : thin gap between axon of one neuron and dendrite of another. Signal exchange Synaptic strength/efficiency
8/10/2019 Introduction ANN
11/19
Introduction
What is an (artificial) neural network A set of nodes (units, neurons, processing elements)
Each node has input and output Each node performs a simple computation by its node
function Weighted connections between nodes
Connectivity gives the structure/architecture of the net
What can be computed by a NN is primarily determinedby the connections and their weights A very much simplified version of networks of neurons
in animal nerve systems
8/10/2019 Introduction ANN
12/19
IntroductionANN
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Nodes
input output node function
Connections connection strength
Bio NN--------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Cell body signal from other neurons firing frequency firing mechanism
Synapses synaptic strength
Highly parallel, simple local computation (at neuron level)achieves global results as emerging property of the interaction
(at network level) Pattern directed (meaning of individual nodes only in the context
of a pattern) Fault-tolerant/graceful degrading Learning/adaptation plays important role.
8/10/2019 Introduction ANN
13/19
Characteristics that can be Extracted
More Complex Network Non Linear Network
Highly Parallel NetworkIt Cannot decide individually, a group decides and fins the solution
Benefits includes Plasticity or Adaptability Not Robust Ability to Learn Different Learning Laws
8/10/2019 Introduction ANN
14/19
Neural Networks
Nervous System
8/10/2019 Introduction ANN
15/19
Properties & Capabilities of NN
Non Linearity Artificial Neuron may be Linear / Non Linear Made up of an interconnection of Non Linear
Neurons Distributed throughout the Network Physical mechanism responsible for the
generation of Input Signal
8/10/2019 Introduction ANN
16/19
Properties & Capabilities of NN
Input Output Mapping Unique Input signal and its Desired responses Minimize the differences between Actual and
Desired Responses Non Parametric estimation with statisitcal
criterion Supervised Learning Paradigm
8/10/2019 Introduction ANN
17/19
Properties & Capabilities of NN
Adaptivity Capability to adapt their synaptic weights to
changes in the environment. Real time synatpic weight change in a non
stationary environment Applications adaptive signal processing, adaptive
control , adaptive pattern classification.
8/10/2019 Introduction ANN
18/19
Properties & Capabilities of NN
Evidential Response Not only which output or pattern but also
confidence in the decision made. Contextual Information
Global activity of all other neurons Fault tolerance
_ Capable of robust Computation
8/10/2019 Introduction ANN
19/19
Properties & Capabilities of NN
VLSI Implementability Uniformity of Analysis & Design
Neurobiological Analogy