1. 1.1 Biological Background
Evolutionary Algorithms in Theory and Practice
1.1 Biological Background
발표자 : 김정집

2. 1.Organic Evolution and Problem Solving
interdisciplinary research field
biology, artificial intelligence, numerical optimization, and decision support
organic evolution
collective learning process within a population of individuals
individual
a search point
container of current knowledge about the “laws” of the environment
fitness value, recombination, mutation, and selection

3. Different Mainstreams
three different Mainstreams
Evolution Strategies (ESs)
Genetic Algorithms (GAs)
Evolutionary Programming(EP)
Index
sec 1.1 biological background
sec 1.2 impact on AI, and ML
sec 1.3 a global optimization algorithm as random search algorithm
sec 1.4 overview of the history of Eas

4. 1.1 Biological Background
Darwinian theory of evolution(Charles Darwin)
natural selection
mutation on phentypes
-> selection under limited environmental conditions
-> advantageous organisms survives

5. Neodarwinism synthetic theory of evolution genes population
transfer units of heredity
changed by mutations
population
evolving unit
consists of a common gene pool
indirect fitness
natural selection as no active driving force
What is mapping from genotype to phenotype?

6. Adaptation
denotes a general advantage in ecological or physiological efficiency
nongenetic-somatic adaptation
genetic adaptation
“To What”
any major kind of environment (adaptive zone)
ecological niche ( the set of possible environments that permit survival of a species)

7. Adaptive surface possible biological trait combinations
natural analogy to the optimization problem
climbing the hill nearest to the starting point
genetic drift
random decrease or increase of biological trait frequencies
dynamically changing by means of environment-population interactions

8. Schematic diagram of an adaptive surface

9. 1.1.1 Life and Information Processing
DNA: 2strands
nucleotide base
Adenine(A), Thymine(T), Cytosine(C), Guanine(G)
purine base (A or G)
pyrimidine base ( T or C)

10. creates the phenotype from the genotype
protein biosynthesis
mapping genotype to phenotype
polygeny - m:1
pleiotropy - 1:m
epistasis
alphabet of amino acids : 20 different one
mRNA(1 strand):transcription,nucleus->ribosomes
tRNA:translation in ribosome

11. the genetic code

12. protein biosynthesis

13. central dogma of molecular genetics
DNA->RNA->Protein
the proof of the incorrectness of Lamarckism

14. Hierarchy of the genetic information

15. 1.1.2 Meiotic Heredity mitosis phylogeny(evolution)
cell division with identical genetic material
phylogeny(evolution)
meiotic cell division

16. Meiosis(I)

17. Meiosis(II)

18. crossover position(s) at random in nature, 1~8 points haploid case(*)
haploid gameter->diploid zygote->haploid cell
recombination and mutaion occur in zygote

19. one-point crossover

20. characteristics of meiosis

21. 1.1.3 mutations
DNA-replication is overwhelmingly exact but not perfect
for a specific gene of the human genome, Pm=6*10-6~8*10-6
by origin
normal-in the replication process
exogenous factors

22. classes of mutations by location usual deviations somatic generative
gene mutations
chromosome mutations
genome mutations

23. gene , genome mutations gene mutations genome mutations
small mutations
little variation-do not negatively effect
large mutations
cause phenotype deviations
progressive(constructive) mutations
cause crossings of boundaries between species
genome mutations
not been tested as an extension of EAs

24. chromosome mutations losses of chromosome regions
deficiencies and deletions
doubling of chromosome regions
duplications
reorganization of chromosomes
translocations and inversions

25. terminal and internal segment losses

26. duplication event

27. inversion event

28. 1.1.4 Molecular Darwinism human genome
consists of one billion nucleotide bases
4^1,000,000,000 possibilities
random emergence of self-reproducing units can be called impossible
explain the efficiency of biological evolution

29. necessary conditions for Darwinian selection
Metabolism
Self-reproduction
Mutation

30. Eigen’s equations
Eigen’s equations for the dynamical behavior of species
: build-up term resulting from self-replication
: term incorporating destruction
: transition probability from class k to I
: growth and shrinking processes of
the total number of individuals

31. Under the assumption of a constant overall organization
buffering the concentrations Ai of energy-rich substances, s.t. AiQi=const
total size of the system is limited
excess productivity
excess productivity must be compensated by transportation through the flow

32. Average excess productivity
Eigen’s eq. Can be transferred to
where , selective value of a species I

33. only those species having Wi above E(t)
will grow the number
shifting E(t) to an optimum
representing Maximum selective value of all species

34. The selection criterion
allow growing of a new species m to become the dominant one
the quasi-species
the currently dominant species together with its stationary distribution of mutants emerging from this species

35. A maximum length s.t. the information can be preserved by reproduction
the ratio of the wild-type(dominant species) reproduction rate to the average reproduction rate of the rest

36. Eigen’s concept of a hypercycle
Experimental results
lmax is no longer than hundred nucleotide bases
Darwinian selection
N=kN
In principle, any new species can grow and become the dominant ones.
Eigen’s concept of a hypercycle
N=kN2
does not allow for diversity of species

37. Summary of experiments
coexistent evolution according to the principle of Darwiniam selection
Hypercyclic system stabilized.
Hypercuclic selection optimizes the system. Only one universal genetic code is produced
The first biological cells emerge
Darwiniam evolution leads to the development of the known variety of species

38. Using a birth and death model, an approximate analytical expression for the dependence of the error threshold
more approximated form