Missing Operators - Githubissues

mlr-org / miesmuschel

Flexible Mixed Integer Evolutionary Strategies

Other

15 stars 4 forks source link

Missing Operators #15

Open mb706 opened 3 years ago

mb706 commented 3 years ago

[ ] sequential operator application
[ ] Intified Gauss
[ ] Uniform Numeric
[ ] Polynomial as in Deb et al. A Combined Genetic Adaptive Search (GeneAS) for Engineering Design
[ ] one-point crossover
[ ] arbitrary-point-crossover
[ ] mean crossover with gaussian (distance-relative)
[ ] Simulated Binary Crossover (Deb et al. Simulated binary crossover for continuous search space. Complex Systems)
[ ] DomHV Selector
[ ] K-tournament Selector (MO)
[ ] Non-dominated sorting selector
[ ] Ranking-base selector (Eiben et al. Introduction to evolutionary computing.)
[ ] Fitness-proportional selector
[ ] mutDoubleGeom
[ ] hamming weight preserving bitflip

mb706 commented 3 years ago

[x] MutatorProxy
[x] RecombinatorProxy
[x] SelectorProxy

mb706 commented 3 years ago

[ ] N-Up recombination operator

sumny commented 3 years ago

Regarding non-dominated sorting: Should we write our own nondom sorting code (most likely in C/C++) oder just reuse some existing implementation?

mb706 commented 3 years ago

I wrote some nondom operations now. Sometimes faster than the C++ versions in ecr :)

sumny commented 3 years ago

thanks! just found the code in the devel branch. some toy benchmark (p = 2, ..., 128 variables; nobs = 10, ..., 1000): reasonably fast I guess!

mb706 commented 3 years ago

hm, it was faster when I tested it, how did you sample your values? There is still one optimization outstanding (doing binary splits when trying out nondominated fronts, instead of going from most to least dominant front), so this should hopefully get better in the next week or so.

(p.s. your y axis should be logscale ;) )

sumny commented 3 years ago

I sampled standard normal distributed values. Did you transpose the data matrix prior to using ecr's version (i.e., each column should contain one point)?