Ev oluationary Computation 1. Computational pro cedures - - PDF document

▶

Feb 24, 2023 127 likes •250 views

Ev oluationary Computation 1. Computational pro cedures patterned after biological ev olution 2. Searc h pro cedure that probabilisti cal l y applies searc h op erators to set of p oin ts in the searc h space

SLIDE 1 Ev

luationary

Computation 1. Computational pro cedures patterned after biological ev

lution

2. Searc h pro cedure that probabilisti cal l y applies searc h

erators to set

ts in the searc h space 169 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997

SLIDE 2 Biological Ev

lution

Lamarc k and

thers:
Sp

ecies \transm ute"

er time Darwin and W allace:

Consisten

t, heritable v ariation among individuals in p

pulation
Natural

selection

the ttest Mendel and genetics:

mec hanism for inheriting traits

genot

yp e ! phenot yp e mapping 170 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997

SLIDE 3 GA(F itness; F itness thr eshol d; p; r ; m)

Initialize:

P p random h yp

theses
Evaluate:

for eac h h in P , compute F itness(h)

While

[max h F itness(h)] < F itness thr eshol d 1. Sele ct: Probabilistic al ly select (1

)p mem b ers

P to add to P S . Pr (h i ) = F itness(h i ) P p j =1 F itness(h j ) 2. Cr

ssover:

Probabilistic al l y select r p 2 pairs

h yp

theses

from P . F

eac h pair, hh 1 ; h 2 i, pro duce t w

spring

b y applying the Crosso v er

erator. Add all

spring

to P s . 3. Mutate: In v ert a randomly selected bit in m

random mem b ers

P s 4. Up date: P P s 5. Evaluate: for eac h h in P , compute F itness(h)

Return

the h yp

thesis

from P that has the highest tness. 171 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997

SLIDE 4 Represen ting Hyp

theses

Represen t (O utl

= O v er cast _ R ain) ^ (W ind = S tr

) b y O utl

W ind 011 10 Represen t IF W ind = S tr

THEN P l ay T ennis = y es b y O utl

W ind P l ay T ennis 111 10 10 172 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997

SLIDE 5 Op erators for Genetic Algorithms

Single-point crossover:

11101001000 00001010101 11111000000 11101010101

Initial strings Crossover Mask Offspring Two-point crossover:

11101001000 00001010101 00111110000 11001011000 10011010011

Uniform crossover: Point mutation:

11101001000 00001010101 10001000100 11101001000 11101011000 00101000101 00001001000 01101011001

173 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997

SLIDE 6 Selecting Most Fit Hyp

theses

Fitness prop

rtionate

selection: Pr(h i ) = F itness(h i ) P p j =1 F itness(h j ) ... can lead to cr

wding

urnamen

t selection:

k h 1 ; h 2 at random with uniform prob.

With

probabilit y p, select the more t. Rank selection:

Sort

all h yp

theses

b y tness

Prob
f

selection is prop

rtional

to rank 174 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997

SLIDE 7 Genetic Programming P

pulation
f

programs represen ted b y trees sin (x) + r x 2 + y

^ sin x y 2 + x +

182 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997

SLIDE 8 Crosso v er

^ sin x y 2 + x + ^ sin x y 2 + x + sin x y + x + ^ sin x y 2 + x + ^ 2

183 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997

SLIDE 9 GP for Classifying Images [T eller and V eloso, 1997] Fitness: based

co v erage and accuracy Represen tatio n:

Primitiv

es include Add, Sub, Mult, Div, Not, Max, Min, Read, W rite, If-Then-Else, Either, Pixel, Least, Most, Av e, V ariance, Dierence, Mini, Library

Mini

refers to a lo cal subroutine that is separately co-ev

Library

refers to a global library subroutine (ev

ed b y selecting the most useful minis) Genetic

erators:

Crosso

v er, m utation

Create

\mating p

and use rank prop

rtionate

repro duction 188 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997

SLIDE 10 Biolog i cal Ev

lution

Lamark (19th cen tury)

Believ

ed individual genetic mak eup w as altered b y lifeti me exp erience

curren t evidence con tradicts this view What is the impact

individual learning

pulation

lution?

189 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997

SLIDE 11 Summary: Ev

lutionary

Program- ming

Conduct

randomized, parallel, hill-cl i m bing searc h through H

Approac

h learning as

ptimization

problem (optimize tness)

Nice

feature: ev aluation

Fitness can b e v ery indirect { consider learning rule set for m ultistep decision making { no issue

assigning credit/blame to indiv. steps 193 lecture slides for textb

Machine L e arning, T. Mitc hell, McGra w Hill, 1997