@@ -157,31 +157,40 @@ sketch - [[file:single-evolve/sketch.pde::void%20populationReset%20u8%20packet][population_reset]] *** experimental results (and playing with the big group) +review the initial [[first_suite][initial experimental results]] -*** next steps +look at the speedup gained through using the big group (with +[[file:scrutinizer.rb][scrutinizer.rb]]) -Let see if it's worthwhile to /meta-evolve/ GP parameters -- central parameter tweaking :: using the ruby script which collects - results and has a central view of the state of the group to - deterministically reset the worst individual in the group at some - rate -- distributed parameter tweaking :: allowing the boards to reset their - neighbors when they feel that they are not up to snuff - -Harder problems (in more sophisticated execution environments than RPN) -- object code :: once generated, would it be possible to burn a .hex - file to a board and execute it in some safe manner? -- simple ant language :: another simple on-board-evaluable language - like RPN with the differences of - - environment :: small 2d grid instead of a number stack - - individual representation :: instead of a string of numbers, a - set of rules which relate states of individual to actions +*** next steps +1) Evaluate the effect of multiple boards on run times + +2) Let see if it's worthwhile to /meta-evolve/ GP parameters + - central parameter tweaking :: using the ruby script which collects + results and has a central view of the state of the group to + deterministically reset the worst individual in the group at some + rate + - distributed parameter tweaking :: allowing the boards to reset their + neighbors when they feel that they are not up to snuff + +3) Harder problems (in more sophisticated execution environments than RPN) + - object code :: once generated, would it be possible to burn a .hex + file to a board and execute it in some safe manner? + - simple ant language :: another simple on-board-evaluable language + like RPN with the differences of + - environment :: small 2d grid instead of a number stack + - individual representation :: instead of a string of numbers, a + set of rules which relate states of individual to actions * Experiments :PROPERTIES: :CUSTOM_ID: experiments :END: ** first suite + :PROPERTIES: + :CUSTOM_ID: first_suite + :END: + 5 runs each of the following goals 1) "xx*" $x^2$ 2) "xxx**" $x^3$ gp4ixm.org ecf45485ee754811de9a685f57c3062ab30a34ca Eric Schulte schulte.eric@gmail.com http://github.com/eschulte/gp4ixm/commit/eb6ebe0a4a26866047d779c910c8d08c2aec3abf eb6ebe0a4a26866047d779c910c8d08c2aec3abf 2009-11-09T11:10:38-08:00 2009-11-09T11:10:38-08:00 fleshing out milestone3 859c5d02ea57f4e335cfe1b15e1c96ddb243e754 Eric Schulte schulte.eric@gmail.com