search

adavoudi / spdnet

Implementation of Deep SPDNet in pytorch
MIT License
58 stars 11 forks source link

Why is gradient orthogonaly projected with 0_n matrix? #12

Open dcwil opened 2 years ago

dcwil commented 2 years ago

In lines: https://github.com/adavoudi/spdnet/blob/2a15e908634cd8db6c75ea45d9e3bd567203eccf/spdnet/optimizer.py#L37-L39

Weight matrix (p.data) is set to 0 before computing the orthogonal projection with it's gradient, why is this the case?

a912289748 commented 1 year ago

而且你这个学习率为什么没有加上