UNIT - IV (Compatibility Mode)

12/9/2012
UNIT- IV
Dr. K. Subrahmanyam Professor in CSE
What we cover in Unit IV

Computer and its role in research, Introduction to evolutionary algorithm Soft Computing and learning in fuzzy systems Fundamentals of Genetic algorithms, Simulated Annealing, Neural Network based optimization, Optimization of fuzzy systems
Computer and its role in research

The computers are indispensable throughout the research process. The role of computer becomes more important when the research is on large sample. The data can be stored in computers for immediate use or can be stored in auxiliary memories. The computer assists researchers throughout the research process.
12/9/2012

Research Phases
Conceptual stage
Search literature, e-databases, etc..
Design & Planning Phase

Softwares for sample size calculation (NCSS-PASS-GESS)
Empirical phase
Data entry, data storage, data management actions, etc..
Analytic phase
Statitistical analysis of data and interpretation ...Calculations using various softwares like (SPSS- STATA-SYSAT)
Dissemination phase
Publishing by converting to pdf and posting on web, etc...
http://www.scribd.com/doc/11515555/Role-of-Computers-in-Research?query=

Through various kinds of analytical software programs, computers are contributing to scientific research in every discipline, ranging from biology to astrophysics, discovering new patterns and providing novel insights.
Computer and its role in research Scientific Simulations

One of the prime uses of computers in pure science and engineering projects is the running of simulations. A simulation is a mathematical modeling of a problem and a virtual study of its possible solutions. Problems which do not yield themselves to experimentation can be studied through simulations carried out on computers.
For example, astrophysicists carry out structure formation simulations, which are aimed at studying how large scale structures like galaxies are formed. Space missions to the Moon, satellite launches and interplanetary missions are first simulated on computers to determine the best path that can be taken by the launch vehicle and spacecraft to reach its destination safely.
http://dualmonitorcard.net/role-of-computers-in-research/
12/9/2012
Introduction to evolutionary algorithm

Evolutionary Algorithms (EA) consist of several heuristics, which are able to solve optimisation tasks by imitating some aspects of natural evolution. They may use different levels of abstraction, but they are always working on whole populations of possible solutions for a given task. EAs are an approved set of heuristics, which are flexible to use and postulate only neglectible requirements on the optimisation task
Heuristic refers to experience-based techniques for problem solving, learning, and discovery. requirements on the optimisation task.
Classification of EA Methods
General EA Process
12/9/2012
Fundamentals of genetic algorithms

GA is a meta heuristic used to find the solution to any combinatorial problem It mimics the mechanism of selection and evaluation. In order to achieve the objective GA generates successive population of alternate solutions until a solution is obtained that yields acceptable results.

Chromosome collection of primitive features called genes. A string of symbols. Population A set of chromosome forms a single population. In GA, a set of strings each represents a potential solution to the given problem. Generation Successive evolution of chromosome is called generation

Offspring- The next generation of chromosome Cross over- Crossover selects genes from the parent chromosomes and create a new offspring. Mutation Mutation is done on the offspring mainly to avoid the solution ending with local optimum. It performs the changing of bits in the offspring from 1 to 0 or from 0 to 1. SelectionIs the process of picking some chromosome from the population for crossover.
12/9/2012
steps of genetic algorithms

Step 1: Input the number of iterations to be carried out. Let the generation count (GC) be 1. Step 2: Generate a random population of n chromosomes ( suitable solutions for problems) Step 3: Evaluate the fitness function f(x) of each chromosome x in the population. Step 4: Create a new population of n chromosomes by repeating the following steps until the new population is complete.

Step 4: Create a new population of n chromosomes by repeating the following steps until the new population is complete.
Step 4.1 : Selection: Sort the population by the objective function value and copy a given percentage (say 20%) of the population for remainder of the population. Then randomly select two parent chromosomes from the above population according to their fitness. Step 4.2 : Crossover : Perform the crossover of the parents for a crossover probability to form a new offspring. If no crossover is performed then the off spring are exact copies of the parents. Step 4.3 : Mutation - perform the mutation of offspring at each locust (position in the chromosome) for a mutation probability. Also evaluate fitness function f(x) of the offspring. Step 4.4 : Acceptance place the new offspring in a new population Step 4.5 : Increment the Generation Count by one ( GC = GC+1)

Step 5 : Use the newly generated population for the next run of the algorithm Step 6: If the end condition is satisfied ( if GC > N) , then go to step 7: otherwise go to step 4. Step 7: Return the best solution in the current population and stop.
12/9/2012
Simulated Annealing
Metropolis et al. (1953) have developed an algorithm to find an equilibrium of a system containing large number of atoms for a known temperature (T). This algorithm forms the basis for the development of many simulated annealing algorithms to solve combinatorial problems.
Simulated Annealing
Step1: Get an initial feasible solution s0 and compute the related value of the objective function f(s0) Step2: Get an initial temperature, T>0. Step3: Generate a feasible solution s1 in the neighbourhood of s0 and compute the related value of the objective function f(s1) Step 4: Compute d = f(s1) - f(s0)
Simulated Annealing
Step5 : if d>= 0 , set s0= s1; otherwise set s0= s1 with the probability e(d/T). Step 6: Set T = r X T. Step 7 : if T > s, then go to step 3; otherwise go to step 8. Step 8: Use local optimization to reach a local optimum starting from the last s0 value.
T is the temperature, r (0 to 1) is used lower the temperature at each computational step and s is the small positive number provided by the user. T, r & s are known as control parameters
12/9/2012
Neural Network based optimization

Introduction to ANN
Neural Network based optimization

Neural-network-based simulation-optimization model for water allocation planning at basin scale
http://www.iwaponline.com/jh/010/0331/0100331.pdf
Neural Network-Based Face Detection Neural network based optimization of drug formulations
http://www.ncbi.nlm.nih.gov/pubmed/12954200
Optimization of fuzzy systems

Fuzzy systems have been successfully applied to problems in classification , modelling, control , and in a considerable number of applications. In most cases, the key for success was the ability of fuzzy systems to incorporate human expert knowledge. In the 1990s, despite the previous successful history, the lack of learning capabilities characterising most of the works in the field generated a certain interest for the study of fuzzy systems with added learning capabilities. Two of the most successful approaches have been the hybridisation attempts made in the framework of soft computing, were different techniques, such as neural and evolutionary, provide fuzzy systems with learning capabilities, as shown in the below figure.
12/9/2012
soft computing and learning in fuzzy systems

Fuzzy Logic
Learning Learning
Evolutionary computation
Neural Networks
soft computing and learning in fuzzy systems

Neuro-fuzzy systems are one of the most successful and visible directions of that effort . A different approach to hybridisation lead to genetic fuzzy systems (GFSs). A GFS is basically a fuzzy system augmented by a learning process based on a genetic algorithm (GA). GAs are search algorithms, based on natural genetics, that provide robust search capabilities in complex spaces, and thereby offer a valid approach to problems requiring efficient and effective search processes. Genetic learning processes cover different levels of complexity according to the structural changes produced by the algorithm , from the simplest case of parameter optimisation to the highest level of complexity of learning the rule set of a rule based system. Parameter optimisation has been the approach utilised to adapt a wide range of different fuzzy systems, as in genetic fuzzy clustering or genetic neuro-fuzzy systems,

UNIT - IV (Compatibility Mode)

Uploaded by

Copyright:

Available Formats

UNIT - IV (Compatibility Mode)

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

UNIT - IV (Compatibility Mode)

Uploaded by

Copyright:

Available Formats

12/9/2012

What we cover in Unit IV

Computer and its role in research

Computer and its role in research

Design & Planning Phase

Computer and its role in research

Computer and its role in research Scientific Simulations

Introduction to evolutionary algorithm

Fundamentals of genetic algorithms

Fundamentals of genetic algorithms

Fundamentals of genetic algorithms

steps of genetic algorithms

steps of genetic algorithms

steps of genetic algorithms

Neural Network based optimization

Neural Network based optimization

Optimization of fuzzy systems

soft computing and learning in fuzzy systems

soft computing and learning in fuzzy systems

You might also like