NGD-SGD icon indicating copy to clipboard operation
NGD-SGD copied to clipboard

Published 20 hours ago • verified publisher icon YiwenShaoStephen
→

Metadata

A Pytorch Implementation of Natural Gradient Descent

Results 1 NGD-SGD issues
Sort by recently updated
recently updated
newest added

optimize.step() takes too long compare to kaldi nnet3 NGD

I guess most time spent on doing the reverse of matrix. But NGD of kaldi nnet3 is fast, why? I have read povey's that paper, have read some code of...

Jarvan-Wang avatar Jarvan-Wang

← Metadata

41
Stars
9
Forks
Watchers

Owner

verified publisher icon YiwenShaoStephen

Metadata

A Pytorch Implementation of Natural Gradient Descent

Back