Vous êtes sur la page 1sur 40

EC 367: Soft Computing

Module 1
Definition of soft computing

Computing is performed by tolerating the


imprecision, uncertainty partial truth and
approximation
Best example for soft computing is human
mind
Fuzzy logic, neural networks, SVM and genetic
algorithms
Soft computing VS Hard computing

Soft computing Hard Computing

Method of approximation Method of accuracy

Methods and data should be precise Need not be precise

Less complex and saves time More complex and time consuming

Generally based on binary logic Based on fuzzy logic


Fuzzy computing

characterizes the uncertainty in engineering


problems
we prefer fuzzy logic for complex system
analysis.
balance precision with uncertainty
Used when fast result is needed
Neural networks

An analogous system to resemble human


brain
It consists of a certain amount of nodes
arranged as several layers
learning through experience.
Genetic algorithms

based on the theory Natural selection

high quality results to optimization and


research problems
Steps in genetic algorithm
Initialization : size of solutions generated
randomly
Selection: each solution is generated randomly
and called generation. In selection a portion of
the solution is chosen for next iteration
Genetic operation: operation on newly selected
solutions. (mutation and cross over)
Termination: above steps are repeated until
conditions are matched.
Applications of soft computing.

Hand writing recognition


Image processing and data compression
Power systems
Fuzzy logic controls
The fuzzy logic. Introduction

Fuzzy logic utilizes the balance between


precision and uncertainty
Achieving a high degree of precision demands
higher computational complexity and time.
Cannot use in high precision applications
Can be used in applications like navigation,
parking, traffic control etc.
In optimizing problems
Consider path optimization for a sales person
visiting cities
But when number is 100 there is 10200
different paths.
In electronic network scenario one similar
situation is to route the signal through nodes
Consider there are 100,000 nodes in a
network;
if we demand 0.75% uncertainty in the
solution, it will take months
If we compromise to 3.5% the solution can be
obtained in some minutes.
Utility of fuzzy systems.
Several sources have shown that fuzzy systems
are universal approximators.
isomorphism between abstract algebra and
linear algebra.
(1)higher degree of complexity and (2)
problems which require approximate solution
in a fast way
Limitations of fussy systems.
shallow model for reasoning

Deep reasoning in which we approximate a


general solution from specific.

Analyzing the data and analyzing the process


Uncertainty and Accuracy.
Closeness to the probability theory.
1. The sample set is fixed in probability theory
2. The frequency of occurrence of an event in
sample space also cannot vary.
3. Outcomes are repeating and totally
independent.
These are not there in fuzzy logic
There are some situations in which probability
cannot be applied.

Process is not purely random

Probability applied on random process.


Uncertainty and Information
Problem associated with precision is the
chance of forgetting.
Majority of material taught in engineering is
of deterministic type.
There will be uncertainty in information
especially associated with a complex system.
Cooking
1. Stop the baking when the color changes to
light brown.
2. Stop when the temperature is exactly 1250C.
Fuzzy sets and membership.
Situation: If a persons height is above 6 feet
or not.

height of a person is 511.9999


1. Analysis by binary logic
2. Analysis by fuzzy logic.
Crisp set satisfies precise properties of
membership.

Fuzzy set satisfies imprecise properties.

Height ranges from 5 to 7 feet is precise but


around 6 is imprecise.
If element x in universe X is to be a member of
set A or not can be expressed mathematically
as.

consider the height example stated above


Its membership function can be represented
as
Classical set and fuzzy set

A classical set is defined by crisp boundaries

A fuzzy set, on the other hand, is prescribed


by vague or ambiguous properties
Classical set or Crisp set.

1)the clock speeds of computer CPUs;


2)the operating currents of an electronic
motor;
3)the operating temperature of a heat pump
(in degrees Celsius);
4)the Richter magnitudes of an earthquake;
the integers 1 to 10.
Operations on crisp sets
Properties of crisp set
Fuzzy Sets
For an element in a universe that contains
fuzzy sets, the transition can be gradual.

varying degrees of membership


in the set.

notation convention for fuzzy sets


notation convention for fuzzy sets

when the universe X, is discrete


When the universe, X, is continuous
Member ship function of fuzzy set
Fuzzy set operations
Union
Itersection
Compliment
Demorgans theorem
Excluded middle & contradiction
axioms
It is different from crisp set
Crisp set middle axioms
Fuzzy set middle axiom

Vous aimerez peut-être aussi