Fluid Evolutionary Algorithms

Embed Size (px)

DESCRIPTION

Presentation of a prototype of evolutionary algorithms based on the FluidDB platform

Citation preview

  • 1. Fluid Evolutionary Algorithms JJ Merelohttp://twitter.com/jjmerelo GeNeura team:http://geneura.wordpress.com University of Granada:http://www.ugr.es

2. Be fluid, my friend FluidDB is an openly writable shared database http://fluidinfo.com/about 3. FluidDB was born from evolutionary computation

  • Company created by Terry Jones.
  • FluidDBis a single web of things.

4. FluidDBmakes it possible for data to be social. 5. Unique yet powerful control model 6. Makes information easy to reuse. 7. How does FluidDB work? Thing fluiddb/about Another Thing http://wcci2010.org jjmerelo/likes dwcorne/stars 5 The sound of music 8. Fluidifying Evolutionary algorithms Thing fluiddb/about Another Thing 0100 1001 1100 1101 jjmerelo/exp-xyz/fitness dwcorne/exp-abc/fitness 33 22 0110 0001 1111 1101 jjmerelo/exp-xyz/current 9. Running a fluid EA 0010101100 001010000 0010101100 0010101100 1110101100 1101101100 0011101100 0010101100 0010101100 1011101100 0011101100 0010101100 10. There is a free... software

  • Free software available fromhttp://sl.ugr.es/001U

11. Implemented in Perl 12. 1600 evaluations 13. Population = 2/32 14. Problem = MaxOnes (64 bits) 15. Number of successful reproductions 16. So far, so good Continuous increase of fitness (steady-state like) 17. Using the pool for improving results This one is first This goes second 18. Decreasing batch size yields better results This one is first This goes second 19. Concluding

  • Concurrency possible, but will have to wait

20. FluidDB can be a persistent pool of problem solutions 21. Balancing using FluidDB and local pool 22. Thanks for your attention Enjoy Barcelona!