[R] GA/SWARM Hyperparameter (HP) Optimisation for Classification based Machine Learning
@gr@ci@n @ending from y@hoo@co@uk
Thu May 3 09:03:57 CEST 2018
I believe that Caret uses a grid-serach approach. I was wondering if:
1 There are more efficient implementations for HP tuning for classification algos (eg XGboost, CatBoost, SVM, RF etc), using say GM/SWARM approaches, akin to Google's approach AutoML for Image related Net problems?
2 This one is most probably wishful thinking, but is anyone looking at GM/SWARM at HP tuning across models (ensemble models). eg the best set of HP for combined XGBoost + SVM, which accounts for the correlation/interaction of the prediction assumptions.
[[alternative HTML version deleted]]
More information about the R-help