search

ddbourgin / numpy-ml

Machine learning, in numpy
https://numpy-ml.readthedocs.io/
GNU General Public License v3.0
15.35k stars 3.72k forks source link

Add the alpha dropout method #24

Closed BoltzmannZhaung closed 5 years ago

BoltzmannZhaung commented 5 years ago

Dear Post Dr.Bourgin, I wanna give a perfect method code, however , unfortunately may be it can not work well (I have not test the method).So.....uh~~ may I ask your help to finish it, then get the honor that be a contributor of this great project (5k+ stars).... THANK YOU!

WuZhuoran commented 5 years ago

Maybe you can refer keras's Alpha Dropout here to create unit test.

And PyTorch Version here.

BoltzmannZhaung commented 5 years ago

How fantastic!!! Thank you~

ddbourgin commented 5 years ago

Hi @BoltzmannZhaung - thanks for doing this, on first glance it looks great :) Will take a closer look soon.

BoltzmannZhaung commented 5 years ago

thank u~!

ddbourgin commented 5 years ago

Hi @BoltzmannZhaung - sorry for the delay. I just did a review pass and left some specific in-line comments. In addition to these, there are two general points I want to emphasize:

  1. Please double check to make sure your code runs when you submit. I noticed at least one bug in your code that would easily have been caught if you tried running it on an example.

  2. It's important that the code in this repo represent a sincere effort at original work. In particular, this means that:

    • It is not directly copied from other sources
    • If it is based off of an existing implementation, those implementations are cited in both the function documentation and the PR itself.

I recognize that for simple functions it can be difficult to implement them cleanly in a way that doesn't resemble existing implementations. This is okay, and I don't expect you to be able to write 100% unique code. What I want, however, is that your code represent your efforts, rather than someone else's :)

BoltzmannZhaung commented 5 years ago

@ddbourgin Thanks for ur review and comments! I will rewrite this method with the individual code immediately that represents my individual efforts. This repo which represents a sincere effort at the original work you have declared did inspire me with determination. THANK YOU! ps: uh... I give you a sincere apology for my previous bad work. for this time, I will try my best to finish it correctly and as original as possible. oh~ also as soon as possible.