Is it possible to change the objective function in fitctree(X,Y,'OptimizeHyperparameters','auto')?

Question

Jan Berthold 2016-11-24

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/313793-is-it-possible-to-change-the-objective-function-in-fitctree-x-y-optimizehyperparameters-auto

回答： Don Mathis 2017-1-8

I would like to use Gmean as object function of hyperparameter optimization. I thought about creating a own kfold loss function with kfoldfun. I don't know if it is possible to pass this the created loss function to 'OptimizeHyperparameters' with a 'gridsearch' or 'randomsearch'. Is this possible in some way or do I have to code a hyperparameteroptimization on my own?

I have already figured it out how to do it with bayesopt. But I would also try a 'gridsearch' and a 'randomsearch'.

Best regards,

Jan

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Don Mathis 2017-1-8

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/313793-is-it-possible-to-change-the-objective-function-in-fitctree-x-y-optimizehyperparameters-auto#answer_249769

You can't change the objective function used to optimize hyperparameters in fitctree.

However, you can do grid search and random search using bayesopt by passing 'AcquisitionFunctionName','grid' or 'AcquisitionFunctionName','random', respectively. Grid search explores the grid in a random order using sampling without replacement.