Matrix-Capsules-EM-PyTorch icon indicating copy to clipboard operation
Matrix-Capsules-EM-PyTorch copied to clipboard

Published 20 hours ago • verified publisher icon yl-1993

Learning Rate

Open tom99763 opened this issue 2 years ago • 0 comments

https://www.kaggle.com/code/tom99763/matrix-capsules-with-em-routing?scriptVersionId=124263074

我發現使用0.001的learning rate沒辦法訓練這個模型(mnist的準確度卡在0.1), 我想知道為什麼訓練這篇paper的模型learning初始要大到0.01, 是gradient的問題嗎?

還有我有發現kernel_tile的運算產生的gradient非常大, 或許是這個問題?

tom99763 avatar Apr 02 '23 04:04 tom99763