search

bajeluk / surrogate-cmaes

Surrogate CMA-ES (S-CMA-ES and DTS-CMA-ES) is a surrogate-based optimizing evolution strategy. It is based on the N. Hansen's CMA-ES algorithm which is interconnected with Gaussian processes (or random forests, that are, however, not maintained here anymore).
19 stars 18 forks source link

Dimensionality reduction #2

Closed tosovvoj closed 9 years ago

tosovvoj commented 9 years ago

new option dimReduction in modelOpts. (Valid only for GP) dimReduction=1 -> reduce at 100% no reduction dimReduction=0.9 -> reduce at 90% etc... All points before training and predictting are trasformed to the basis of BD and then is used only first dimReduction dimensions of points for train and predict.

tosovvoj commented 9 years ago

Ahoj Lukáši, je to můj první pull-reques a nejsem si uplně jistý, že jsem vše udělal git-ok, dáš mi prosím vědět, ať už se request povede nebo ne :)? Dík Vojta