lsq-net icon indicating copy to clipboard operation
lsq-net copied to clipboard

Published 20 hours ago • verified publisher icon zhutmost

Train and validation set are imbalanced

Open zhutmost opened this issue 4 years ago • 2 comments

When using the random_split function provided in PyTorch to split the dataset to training and validation sets, the propotion of categories are imbalanced.

zhutmost avatar Jun 05 '20 08:06 zhutmost

Fixed in commit https://github.com/zhutmost/lsq-net/commit/02f802e6652d3d067b953f756f01dd2d50d54456.

zhutmost avatar Jun 06 '20 02:06 zhutmost

However, the imbalance among categories still exists. Maybe need another Sampler.

zhutmost avatar Jun 06 '20 02:06 zhutmost