Download Neural Nets in Forecasting

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Neural Networks
And
Its Applications
By
Dr. Surya Chitra
OUTLINE
• Introduction & Software
• Basic Neural Network & Processing
– Software Exercise Problem/Project
• Complementary Technologies
– Genetic Algorithms
– Fuzzy Logic
• Examples of Applications
– Manufacturing
– R&D
– Sales & Marketing
– Financial
OPTIMIZATION PROBLEM
• Combination of variables which
produces the best results
• When No. of variables increase,
it is difficult to do optimization.
• Genetic Algorithms serve as an
alternative to optimization in
feature selection
BIOLOGICAL EVOLUTION
• Evolution of species - “Struggle for Life”
• Best individuals have greatest
probability of surviving and winning
battles for reproduction.
• When a combination of two good
genomes generates better genetic
material.
• GAs are inspired by evolution theory.
Definition of Genetic
Algorithms
Genetic Algorithms are Search
Algorithms Based on the Mechanics of
Natural Selection and Natural Genetics
Goldberg (1989)
Genetic Algorithms are Software,
Procedures Modeled After
Genetics and Evolution
Bauer (1993)
Genetics & Genetic Algorithms
HUMAN CELL
GENETIC ALGORITHMS
CHOMOSOMES
STRING STRUCTURES
(Dictate Hereditary of Individual)
(Strings are Rated by Fitness
Function)
INDIVIDUAL GENES
(Encodes Specific Feature
Actual Value is Called Allele)
23 Chromosomes
23 Chromosomes
PARENTS
ELEMENTS in STRINGS
Actual Value Stored in Elements)
Crossover of strands
Mutation of strands
(diversification)
New Strand of
Chromosomes
OFFSPRING
Basis for Genetic Algorithms
• Randomized Search
– Strings are Chosen & Combined
Stochastically
• Based on Survival of the Fittest
– Uses Fitness Function
• Select Fittest String to Create New String
• Based on Interbreeding Population
– To Create Innovative Search Strategy
EVOLUTIONS IN GA
Experimental conditions leading to better
results will prevail over the worst ones
and an improvement can be obtained
by a recombination with some random
changes.
Experimental Conditions -----> Genome
Variables ---------> Genes
Response ----------> Measure of fitness
BASIC STEPS IN GAs
• Coding of Variables
• Initiation of population
• Evaluation of the response
– Reproduction
– Crossover
– Mutations
Steps 3 to 6 alternate until a termination criterion is reached.
(Lack of improvement or maximum number of generations)
Reproduction
• Reproduction allows individual
strings to be copied for next
generation.
• The chance that a string is copied
depends on its fitness function.
01110
10000
01001
STRING
9
12
5
FITNESS VALUE
35%
46%
19%
PERCENTAGE
Crossover
• Biologically it is the blending of
Chromosomes.
• Selects two strings at random &
calculates whether crossover
should take place based on
crossover probability.
MUTATIONS
• Mutations are irregular changes
with very low probability of
occurrence and affect single
gene.
• Sometimes mutations generate
good results and contribute to
evolution.
Genetic Algorithm Iteration Loop
Example of GA Optimization
Algorithm
Assume: Answer is Integer and between 0 and 25
STRING
00001
00101
10110
DECODED VALUE
1
5
22
Example of GA Optimization
Algorithm
Example
STRING
00101
01101
10110
X VALUE
5
13
22
F(X)
25
169
484
RELATIVE FITNESS
0.04
0.25
0.71
Example of GA Optimization
Algorithm
First Iteration
STRING POPULATION
00010
00111
10110
01011
Selection
STRING
X VALUE
F(X)
00010
00111
10110
01011
2
7
22
11
27
22
-293
-18
RELATIVE
FITNESS
0.35
0.34
0.008
0.30
NUMBER OF
SELECTIONS
1
2
0
1
Example of GA Optimization
Algorithm
Mating Pool
STRING POPULATION
00111
00010
00111
01011
Crossover
MATING POOL
STRINGS
0001|0
0011|1
01|011
00|010
NEW POPULATION
00011
00110
01010
00011
Example of GA Optimization
Algorithm
End of First Iteration and New Population
STRING POPULATION
00011
00110
01010
00011
X VALUE
3
6
10
3
Definition of Fuzzy Logic
Systems
Fuzzy Logic is a multi-valued logic that
allows intermediate values to be defined
between conventional evaluations like
yes/no, true/false, etc.
More Human-like way of thinking in
the programming of computers.
Initiated by Lotfi Zadeh
Computer Science
Univ. of California at Berkeley
Fuzzy Sets
A = [ 5, 8 ]
Fuzzy Sets
B = [ Set of Young People ]
B = [ 0, 20 ]
Operations on Fuzzy Sets
Fuzzy Set between 5 and 8
Fuzzy Set about 4
Operations on Fuzzy Sets
Fuzzy Set between 5 and 8 AND about 4
Operations on Fuzzy Sets
Fuzzy Set between 5 and 8 OR about 4
Operations on Fuzzy Sets
Negation of Fuzzy Set between 5 and 8
Related documents