RektPunk comments

Results 14 comments of


                                            RektPunk

Multiple quantile regression with preserving monotonicity

Hey, @juandados. Thanks for asking :) As you said, hessian should be null, but for computation, hess is set as 1. Also, the first order approximation (cause only use grad)...

Multiple quantile regression with preserving monotonicity

Hey @StatMixedML, thxs for your comment. The first comment is exactly what i concerned. I also wonder is there any way to remove redundancy :). For second comment, I don’t...

Multiple quantile regression with preserving monotonicity

Hey, thanks for conducting experiment. I think hyper-parameters should be different cause constraint model is much complex and also have constraint as you mentioned.

Multiple quantile regression with preserving monotonicity

Yeah, I totally agree with you. For that case, other trick should be considered.

Multiple quantile regression with preserving monotonicity

LightGBM version has changed and some details have been modified. The latest code can be found [here](https://github.com/RektPunk/quantile-tree). I also added code for xgboost.

03-01-01 Definition · 모두를 위한 컨벡스 최적화

맞아요, 수정 pr 올렸습니다 :)

01-02 Convex optimization problem · 모두를 위한 컨벡스 최적화

늦었지만 해당 부분 증명을 좀 더 직관적으로 변경했어요.

Base64Bytes field doesn't raises validation error for min_length constraint

I think this is related to encode, decode logic since even when converted `Base64Bytes` to `Base64Str`, there is no validation error too. We can avoid it by creating a custom...

Missing Dependency: `matplotlib.pyplot`

Thanks @Ankush1oo8, but it's in https://github.com/microsoft/TinyTroupe/pull/33 already. please close https://github.com/microsoft/TinyTroupe/pull/60

Enhance `optuna/optuna/visualization` by eliminating redundant for-loops and repetitive code

Can I give this a try?