Exercise 1
description
Transcript of Exercise 1
Exercise 1
Francesco AbateNiccolo` BattezzatiMiguel Kaouk
Apprendimento mimetico
EP – Program Flow
Generate first population
Generate new population by
mutation
Selection by tournament
Goal?
Max iterations?
END
μ
q
σ, c
.MAX_ITER
μ + μ
YES
YES
NO
NO
Fitness evaluation
EP – Program Architecture
ep.confEvoConfigParser
EvoConfigurator
main
EvoAgent
( float x[D] )
float evaluate_fitness(float (*fitnessFnc)(EvoAgent *))bool termination(bool (*terminationFnc)(EvoAgent *))
Experimental results
μ qmean # of fitness evaluations
σ = 0.8 σ = 1.0 σ = 2.5
10 10 n.s. 237946 379698
10010 n.s. 227399 560768
50 n.s. 208810 1712028
D = 2, static σ
Experimental results
μmean # of fitness evaluations
q = 10 q = 50
10 3050 /
100 18633 30033
1000 61666 56000
D = 2, dynamic σ
Experimental results
μmean # of fitness
evaluationsmean # of
generations
q = 10 q = 50
10 50847 5084 / /
100 264310 2642 310920 3109
1000 2008700 2008 1661200 1661
D = 5, dynamic σ
Experimental results
μmean # of fitness evaluations
q = 10 q = 50
10 140058 /
100 851783 823050
1000 5567500 6165000
D = 10, dynamic σ