Title: Gene Expression Programming: A New Adaptive Algorithm for Solving Problems
1Gene Expression Programming A New Adaptive
Algorithm for Solving Problems
- Vincent Chan
- System Electronics Lab.
- Seoul National University
2Contents
- Introduction
- Overview of Gene Expression Algorithm
- The genome of GEP individuals
- Fitness functions and selection
- Reproduction with modification
- Example
- Conclusion
3Introduction to GEP
- Similarities with GA,GP
- Using populations of individuals
- Seleting individuals accoding to fitness
- Introduces genetic variation using
- Genetic Operatrs.
- Solving searching, sorting and etc. problems
4Introduction to GEP
- Difference among GEP,GA,GP resides in the nature
of the individuals - Individuals
- GA Linear strings of fixed length/Array of bits
- GP Nonlinear entities of different sizes and
shapes (Parse trees)/Tree-like representations - GEP
- Encoded as linear strings of fixed length
- Genome or chromosomes
- Expressed as nonlinear entities of different size
and shapes - Simple diagram representations or expression
trees
5Introduction to GEP
- Advantages of GEP
- Truly functional genotype/phenotype relationship
- Enabled by GEP chromosomes organization
- Translation from language of chromosomes to
languages of ET(Expression trees) - Translation from genotype to phenotype
6Overview of GEA
Create Chromosomes of initial Population
Replication
Mutation
IS transposition
Express Chromosomes
RIS transposition
Evaluate Fitness
1-point Recombination
Iterate or Terminate
End
2-point Recombination
Gene Recombination
Keep Best Program
Prepare New Programs of Next Generation
Select Programs
7The genome of GEP individuals
- Open reading frames and genes
- Gene expression programming genes
- Multigenic chromosomes
- Expression trees and the phenotype
- Information decoding translation
- Interactins of sub-expression trees
8Open Reading frames and genes
- Structural Organization of GEP
- Apply the biology idea
- Gene begins with Start codon
- Continues with amino acid codons
- Ends at termination codons
9Codon sequence and codon table
10Example of ORFs
Algebraic Expression
Expression Tree (ET)
Genotype
11GEP genes
- A head and a tail
- Head
- Contains symbols representing both functions and
terminals - Tail
- Contains only terminals
- Function
- Elements from function set F
- Terminal
- Elements from the terminal set T
12GEP genes
Tail
Head
Expression Tree (ET)
13Multigenic chromosomes
- More than one gene of equal length in GEP
chromosomes - Sub-Ets
14Expression trees and the phenotype
- Information decoding translation
- Interactons of sub-expression trees
15Information decoding Translation
- Expression of geneti information starts from
translation - Transfer from a gene into an ET
- In GEP gene--gtET is done directly
- Single gene organism of gene
- GEP chromosomes are composed of one or more ORFs
- Single ET Multi-subunit ET
- The whole organism is encoded in a linear genome.
16Interactions of sub-expression trees
A Two-genic chromosome
Sub-Ets codified by each gene
The result of posttranslational linking
17Fitness functions
- Fitness functions
- Symbolic regression/function finding is an
improtant application of GEP - To find an expression that performs well for all
fitness cases within a certain eror of the
correct value. - Limit the range of selection to control the
population of the result - How to decide the range/limitation?
18Fitness functions
- Best possible solution
- With in a minimum error
- A board limit for selection (20)
- For evolutionary to get started
- Then it could be reshaped by seletion and
populations adaption - Finding better solutions --gt perfect solution
19Selection (Same as GA)
- Selecting individuals
- According to fitness
- By roulettewheel sampling Goldberg 1989
- Or tournament selection
- No scheme is perfect for any problem
- Conclusion from the paper
- For more complex problems it seems that
roulettewheel selection with elitism is best
20Reproduction with modification
- Using result from fitness and selection
- Reproduce with modification
- Replication
- Mutation
- Transpositon and insertion sequence elements
- Transpositon of insertion sequence elements
- Root transposition
- Recombination
- One-point recombination
- Two-point recombination
21Examples of GEP in problem solving
- Symbolic regression
- Sequence inducton and th creation of constants
- Block stacking
- Evolving cellular automata rules for the
density-classification problem - Boolean concept learing
22Conclusions
- Detailed implementation of GEP is explained
- It could be applied in several different fields
with advantage of running efficiently in a PC - GEP offers more opportunities for soloving more
complex problems - Organization of chromosomes is originally
introduced - GEP could present nature more faithfully(?), so
it could be used as computer models of natural
evolutionary processes.
23- Thank You!
- ?????!
- ????!
- ???!
- Danke schön!
- C?m on!