Scribd
Upload
36 views

Soft Hard

This document provides an overview of soft computing techniques. It introduces artificial neural networks (ANN), fuzzy logic, and genetic algorithms. ANN works best for non-linear relationships and learns through experience. Fuzzy logic uses partial set membership to model imprecise concepts. Genetic algorithms are inspired by natural selection. The techniques are complementary and find solutions to complex problems by exploiting uncertainty. Examples of applications include pattern recognition, control systems, prediction, and more.

Uploaded by

Pjjhag
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
36 views

Soft Hard

This document provides an overview of soft computing techniques. It introduces artificial neural networks (ANN), fuzzy logic, and genetic algorithms. ANN works best for non-linear relationships and learns through experience. Fuzzy logic uses partial set membership to model imprecise concepts. Genetic algorithms are inspired by natural selection. The techniques are complementary and find solutions to complex problems by exploiting uncertainty. Examples of applications include pattern recognition, control systems, prediction, and more.

Uploaded by

Pjjhag
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 14

Unit 1

Lecture 2
Topics
• Review
• Major areas of soft computing
• Brief introduction to ANN
• Brief introduction to Fuzzy Logic
• Brief introduction to Genetic Algorithms
• Applications of soft computing techniques
Hard Computing Vs Soft Computing
Brief introduction to major areas of Soft Computing
• Soft computing is a combination of several methods
• Artificial Neural Networks(ANN) or Neural Net
• Fuzzy Logic
• Genetic Algorithms
• All these methods are complimentary to each other and can be used to solve
a complex problem
• Exploits the imprecision and uncertainty in a decision making process
1) Artificial Neural Networks(ANN)
• An information processing model inspired by biological neuron system
• Works best when the relationship between input and output is highly non
linear
• Has a set of neurons and each neuron performs some operation
• Overall operation of the network is the weighted sum of these operations
• Consists of three layers: input, hidden, output
• Each layer can have any no of nodes
• Fast response due to their parallel architecture
• Well suited for real time systems
• X1 and X2 are the input neurons
• Transmits signals to the output neuron Y which receives the signal
• Input neurons connect to output through weighted interconnection links with
weights W1 and W2
• Output = function(net input calculated)
=f(x1w1+x2w2)
2) Fuzzy logic
• Partial set membership than crisp set membership(1 or 0) or non membership
• Notions like rather warm, pretty cold etc can be formulated mathematically in
fuzzy logic
• Fuzzification process : Crisp input values to a fuzzy system are converted into
a fuzzy value
• Defuzzification process : produces crisp outputs
• Fuzzification of crisp input is done by the fuzzification interface
• Interface engine takes the fuzzy value generated and the fuzzy rule base (IF--
THEN) and generates fuzzy outputs
• Defuzzification interface generates the crisp output actions from the fuzzy
output
3) Genetic Algorithms
• Inspired by nature
• Basic processes:
• Initialization: initial population created randomly
• Evaluation: Each member of the population is evaluated for fitness
• Selection: The ones that fits the desired requirements are selected
• Cross over: new members created by combining the best aspects of the existing
members
Hybrid Systems
• Neuro fuzzy hybrid systems
• Neuro genetic hybrid systems
• Topology optimization done for neural networks using the genetic algorithms
• Fuzzy genetic hybrid system
• Genetic algorithm used to develop the best set of rules for the rule base
Neural Networks: important for their ability to adapt and learn
Fuzzy logic: important for its exploitation of partial truth and imprecision
Genetic Algorithms: important for its optimization
Probability reasoning: control of randomness and uncertainty
• Constituents of soft computing are not competitive but complementary
• Offers their own advantages and techniques and find solutions to otherwise
insolvable problems
Applications:
• Aerospace
• Aircraft and air traffic control using neural networks
• Automated using inputs as location, altitude, direction, speed of each radar blip
• Output is Air traffic controller’s instruction
• Space craft control
• Communication systems
• Data communication
• Neural n/w used for data compression
• Communication networks
• Multimedia traffic prediction using neuro fuzzy systems
• Consumer appliances
• Neural network based air conditioners
• Dishwashers, washing machines,microwave ovens etc based on neuro fuzzy systems
• Manufacturing Automation
• Robotics
• Robotic movement
• Robotic arms
• Transportation
• Elevator controls based on neuro fuzzy systems
• Traffic control
• Pattern recognition
• Handwritten digits recognition
• Voice recognition
• Photos and fingerprint recognition
• Medicine
• Medical diagnosis
• Tissue guided surgery
•Molecular Biology
•For gene prediction
• Music composition
• Neural networks used for predicting notes in music
• Fuzzy systems used for studying music emotions
• Evaluation of compositions using various techniques
• Weather prediction
Conclusion
• Hard computing vs soft computing a diagrammatic representation
• Major areas of soft computing
• Introduction to ANN
• Introduction to Fuzzy Logic
• Introduction to Genetic Algorithms
• Some applications of soft computing techniques

You might also like