Even better Mastermind

18
J. J. Merelo, Carlos Cotta, Antonio Mora U. Granada & Málaga (Spain) Http://geneura.wordpress.com http://twitter.com/geneura Optimizing worst-case scenario in evolutionary solutions to the MasterMind puzzle

Transcript of Even better Mastermind

Page 1: Even better Mastermind

J. J. Merelo, Carlos Cotta, Antonio Mora

U. Granada & Málaga (Spain)

Http://geneura.wordpress.com

http://twitter.com/geneura

Optimizing worst-case scenario in evolutionary

solutions to the MasterMind puzzle

Page 2: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 2

Game of MasterMind

Page 3: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 3

7 reasons why you should care

Donald Knuth

NP-Complete

Differential cryptanalisis/ATM cracking

Circuit/program test

Genetic profiling

Optimal solution not known

Interesting search problem

Page 4: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 4

Let's play, then

Page 5: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 5

Consistent combinations

Page 6: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 6

Naïve Algorithm

Repeat

Find a consistent combination and play it.

Page 7: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 7

Looking for consistent solutions

Optimization algorithm based on distance to consistency (for all combinations played)

D = 2

Page 8: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 8

Not all consistent combinations are born the same

There's at least one better than the others (the solution).

Some will reduce the remaining search space more.

But scoring them is an open issue.

Page 9: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 9

What we did beforeWhat we did before

Apply heuristic methods to speed up finishing Apply heuristic methods to speed up finishing gamesgames

Page 10: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 10

What we do now

Increase diversity in search via new operators and selection mechanisms

Page 11: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 11

Objectives

Reduce the probability of takeover by a single individual

Reduce the possibility of repeated generation of a single combination

Increase speed to afford tackling bigger sizes

Page 12: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 12

New tricks for old games

Add permutation operator

Add diff uniform crossover

crossover over different positions only

Selection makes sure parents are different

Change to tournament selection

Stronger selective pressure

Higher replacement rate

Page 13: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 13

Results: number of evaluations

Page 14: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 14

Results: any good at MasterMind?

Better

Bet

ter

Page 15: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 15

Mission accomplished?

Page 16: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 16

Worst case is better!

Measures to increase diversity have a positive impact on quality of algorithm and also speed

Page 17: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 17

Open source your science!

Page 18: Even better Mastermind

Even B etter Evolutionary Mastermind - Merelo/Cotta/Mora 18

Thank you very much

Questions?