ymattu
commented
6 years ago I couldn't reproduce the error, as following.
> set.seed(71)
> res0 <- xgb_cv_opt(data = df_example,
+ label = y,
+ objectfun = "binary:logistic",
+ evalmetric = "auc",
+ n_folds = 3,
+ classes = 10,
+ init_points = 4,
+ n_iter = 5)
elapsed = 0.01 Round = 1 eta_opt = 0.4475 max_depth_opt = 5.0000 nrounds_opt = 86.7596 subsample_opt = 0.1971 bytree_opt = 0.7882 Value = 0.5852
elapsed = 0.01 Round = 2 eta_opt = 0.1121 max_depth_opt = 5.0000 nrounds_opt = 144.4636 subsample_opt = 0.7513 bytree_opt = 0.8698 Value = 0.6113
elapsed = 0.01 Round = 3 eta_opt = 0.4441 max_depth_opt = 5.0000 nrounds_opt = 130.1620 subsample_opt = 0.4701 bytree_opt = 0.7318 Value = 0.5542
elapsed = 0.01 Round = 4 eta_opt = 0.8827 max_depth_opt = 5.0000 nrounds_opt = 141.4816 subsample_opt = 0.8389 bytree_opt = 0.7178 Value = 0.5254
elapsed = 0.01 Round = 5 eta_opt = 0.1328 max_depth_opt = 5.0000 nrounds_opt = 104.9175 subsample_opt = 0.3617 bytree_opt = 0.8559 Value = 0.5064
elapsed = 0.01 Round = 6 eta_opt = 0.8979 max_depth_opt = 5.0000 nrounds_opt = 153.8820 subsample_opt = 0.1759 bytree_opt = 0.8697 Value = 0.5993
elapsed = 0.01 Round = 7 eta_opt = 0.4374 max_depth_opt = 5.0000 nrounds_opt = 141.3386 subsample_opt = 0.1684 bytree_opt = 0.8699 Value = 0.3771
elapsed = 0.01 Round = 8 eta_opt = 0.7030 max_depth_opt = 4.0000 nrounds_opt = 144.5133 subsample_opt = 0.4369 bytree_opt = 0.8664 Value = 0.5131
elapsed = 0.02 Round = 9 eta_opt = 0.6536 max_depth_opt = 6.0000 nrounds_opt = 147.7658 subsample_opt = 0.7710 bytree_opt = 0.8733 Value = 0.6066
Best Parameters Found:
Round = 2 eta_opt = 0.1121 max_depth_opt = 5.0000 nrounds_opt = 144.4636 subsample_opt = 0.7513 bytree_opt = 0.8698 Value = 0.6113 I wonder that error is because of random seed or something...
If you still get error, try to increase the kappa parameter of acquisition function.
Infinite values of the Deviance Function
This error often means the searching parameter fails because of local maximum point.
Increasing kappa enables to enlarge the parameter space to search.
ghost
TheXu
orrymr
When using the
xgb_cv_optfunction I get the following error after four rounds.Error in GP_deviance(param_init_200d[i, ], X, Y, nug_thres, corr = corr) : Infinite values of the Deviance Function, unable to find optimum parametersI found this page
https://github.com/yanyachen/rBayesianOptimization/issues/36
talking about the same error in the rbayesianoptimization package and tried playing around with the parameter ranges but still cannot get the optimization to run. Reproducible example below.