WisconsinAIVision / W2D

The Two Dimensions of Worst-case Training and the Integrated Effect for Out-of-domain Generalization
8 stars 2 forks source link

Questions about code of "W2D CMNIST only" #1

Open judgingalready opened 2 years ago

judgingalready commented 2 years ago

Hi , excuse me.

I am interested about the special training of Colored MNIST in your paper, which gets great performance. However, I find that something about biased classifier did not appear in the code about "W2D CMNIST only". I'm not sure if it's my mistake, and hope to get more details about it.

Thanks a lot.

OoDBag commented 2 years ago

Hi, thanks for your reminder. We omit it as it's not the main focus of our paper but we will add the modified W2D implementation shortly. In case you need details now, I have attached a previous slide of modified W2D on CMNIST below for reference.

Screen Shot 2022-09-30 at 6 34 43 PM
judgingalready commented 2 years ago

Thanks for your quick reply and details !