1 / 14

Population Sizing as an Emergent Behavior

Population Sizing as an Emergent Behavior. Jason Cook. Motivation. Ease of use Limit necessary manual tuning Potentially improve performance. Problem Statement. Remove the population size parameter Introduce no new parameters Maintain useful level of accuracy. Solution Method.

fayola
Download Presentation

Population Sizing as an Emergent Behavior

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Population Sizing as an Emergent Behavior Jason Cook

  2. Motivation • Ease of use • Limit necessary manual tuning • Potentially improve performance

  3. Problem Statement • Remove the population size parameter • Introduce no new parameters • Maintain useful level of accuracy

  4. Solution Method • Set up the population size as an emergent behavior • Replace survivor selection with a survival chance • SRi = (Fi – Fmin) / (Fmax – Fmin)

  5. Experimental Setup • Compare with a traditional EA • Two main test problems: Griewank Function and D-TRAP Problem • Survival Method: • Survival Chance or • N-Tournament • Other operators and parameter held constant for both EAs

  6. Griewank Function • N dimensional minimization problem • Many local optima • Optimal solution: xi = 0

  7. D-TRAP • 250 4-bit pieces • Each piece is worth: • 3 – u if u ≤ 3 • 4 otherwise • Optimal solution: xi = 0

  8. Results

  9. Results (Continued):

  10. Results (Continued):

  11. Results (Continued):

  12. Analysis • Performs as well or better than a traditional EA • Still affected by the initial population size

  13. Future Work • More test problems • Use a more competitive EA to test against

  14. Questions?

More Related