Evolutionary Programming

Chapter 5
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

EP quick overview
 Developed: USA in the 1960’s
 Early names: D. Fogel
 Typically applied to:
– traditional EP: machine learning tasks by finite state machines
– contemporary EP: (numerical) optimization
 Attributed features:
– very open framework: any representation and mutation op’s OK
– crossbred with ES (contemporary EP)
– consequently: hard to say what “standard” EP is
 Special:
– no recombination
– self-adaptation of parameters standard (contemporary EP)
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

EP technical summary tableau

Representation Real-valued vectors

Recombination None

Mutation Gaussian perturbation

Parent selection Deterministic

Survivor selection Probabilistic (+)

Specialty Self-adaptation of mutation

step sizes (in meta-EP)
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Historical EP perspective

 EP aimed at achieving intelligence

 Intelligence was viewed as adaptive behaviour
 Prediction of the environment was considered
a prerequisite to adaptive behaviour
 Thus: capability to predict is key to intelligence
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Prediction by finite state machines

 Finite state machine (FSM):

– States S
– Inputs I
– Outputs O
– Transition function  : S x I  S x O
– Transforms input stream into output stream
 Can be used for predictions, e.g. to predict
next input symbol in a sequence
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

FSM example

 Consider the FSM with:

– S = {A, B, C}
– I = {0, 1}
– O = {a, b, c}
  given by a diagram
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

FSM as predictor

 Consider the following FSM

 Task: predict next input
 Quality: % of in(i+1) = outi
 Given initial state C
 Input sequence 011101
 Leads to output 110111
 Quality: 3 out of 5
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Introductory example:
evolving FSMs to predict primes
 P(n) = 1 if n is prime, 0 otherwise
 I = N = {1,2,3,…, n, …}
 O = {0,1}
 Correct prediction: outi= P(in(i+1))
 Fitness function:
– 1 point for correct prediction of next input
– 0 point for incorrect prediction
– Penalty for “too much” states
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Introductory example:
evolving FSMs to predict primes
 Parent selection: each FSM is mutated once
 Mutation operators (one selected randomly):
– Change an output symbol
– Change a state transition (i.e. redirect edge)
– Add a state
– Delete a state
– Change the initial state
 Survivor selection: (+)
 Results: overfitting, after 202 inputs best FSM had one
state and both outputs were 0, i.e., it always predicted “not
prime”
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Modern EP

 No predefined representation in general

 Thus: no predefined mutation (must match
representation)
 Often applies self-adaptation of mutation
parameters
 In the sequel we present one EP variant, not
the canonical EP
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Representation

 For continuous parameter optimisation

 Chromosomes consist of two parts:
– Object variables: x1,…,xn
– Mutation step sizes: 1,…,n
 Full size:  x1,…,xn, 1,…,n 
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Mutation
 Chromosomes:  x1,…,xn, 1,…,n 
 i’ = i • (1 +  • N(0,1))
 x’i = xi + i’ • Ni(0,1)
   0.2
 boundary rule: ’ < 0  ’ = 0
 Other variants proposed & tried:
– Lognormal scheme as in ES
– Using variance instead of standard deviation
– Mutate -last
– Other distributions, e.g, Cauchy instead of Gaussian
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Recombination

 None
 Rationale: one point in the search space
stands for a species, not for an individual and
there can be no crossover between species
 Much historical debate “mutation vs. crossover”
 Pragmatic approach seems to prevail today
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Parent selection

 Each individual creates one child by mutation

 Thus:
– Deterministic
– Not biased by fitness
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Survivor selection
 P(t):  parents, P’(t):  offspring
 Pairwise competitions in round-robin format:
– Each solution x from P(t)  P’(t) is evaluated against q
other randomly chosen solutions
– For each comparison, a "win" is assigned if x is better
than its opponent
– The  solutions with the greatest number of wins are
retained to be parents of the next generation
 Parameter q allows tuning selection pressure
 Typically q = 10
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Example application:
the Ackley function (Bäck et al ’93)
 The Ackley function (here used with n =30):
 1 n 2   1 n 

f ( x)  20  exp  0.2   xi   exp  cos(2xi )   20  e

 n i 1   n i 1 
 Representation:
– -30 < xi < 30 (coincidence of 30’s!)
– 30 variances as step sizes
 Mutation with changing object variables first !
 Population size  = 200, selection with q = 10
 Termination : after 200000 fitness evaluations
 Results: average best solution is 1.4 • 10 –2
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Example application:
the Ackley function (Bäck et al ’93)
 The Ackley function (here used with n =30):
 1 n 2   1 n 

f ( x)  20  exp  0.2   xi   exp  cos(2xi )   20  e

 n i 1   n i 1 
 Representation:
– -30 < xi < 30 (coincidence of 30’s!)
– 30 variances as step sizes
 Mutation with changing object variables first !
 Population size  = 200, selection with q = 10
 Termination : after 200000 fitness evaluations
 Results: average best solution is 1.4 • 10 –2
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Example application:
evolving checkers players (Fogel’02)

 Neural nets for evaluating future values of moves are

evolved
 NNs have fixed structure with 5046 weights, these are
evolved + one weight for “kings”
 Representation:
– vector of 5046 real numbers for object variables (weights)
– vector of 5046 real numbers for ‘s
 Mutation:
– Gaussian, lognormal scheme with -first
– Plus special mechanism for the kings’ weight
 Population size 15
 A.E. Eiben and J.E. Smith, Introduction to Evolutionary Computing
Evolutionary Programming

Example application:
evolving checkers players (Fogel’02)

 Tournament size q = 5
 Programs (with NN inside) play against other
programs, no human trainer or hard-wired
intelligence
 After 840 generation (6 months!) best strategy
was tested against humans via Internet
 Program earned “expert class” ranking
outperforming 99.61% of all rated players