- Implement differential evolution
- Implement CMAES
- Implement Particle Swarm Optimization
- Improve tournament selection testing
- Refactor models testings
- Add more context to errors (at least the method/struct name) (https://dave.cheney.net/2016/04/27/dont-just-check-errors-handle-them-gracefully)
- Implement operators described in http://www.ppgia.pucpr.br/~alceu/mestrado/aula3/IJBB-41.pdf
- http://deap.readthedocs.io/en/master/
- http://pyevolve.sourceforge.net/intro.html#ga-features
- http://www.dmi.unict.it/mpavone/nc-cs/materiale/moscato89.pdf
- Keep names short
- Aim for 100 characters per line
// Bad
x := 42
// Good
var x = 42Please inspire yourself from the existing algorithms before implementing, the naming conventions are easy to grasp.
Genetic algorithms are notorious for being embarrassingly parallel. Indeed, most calculations can be run in parallel because they only affect part of the GA. Luckily Go provides good support for parallelism. As some gophers may have encountered, the math/rand module can be problematic because there is a global lock attached to the random number generator. The problem is described in this StackOverflow post. This can be circumvented by providing each population with it's own random number generator.
go test -bench . -cpuprofile=cpu.profgo tool pprof -pdf eaopt.test cpu.prof > profile.pdf