search

Eric-mingjie / network-slimming

Network Slimming (Pytorch) (ICCV 2017)
MIT License
907 stars 214 forks source link

Sparse Confusion #76

Closed EddieEduardo closed 2 years ago

EddieEduardo commented 2 years ago

Thanks for sharing the codes! I am confused about the L1 referred in the paper.

Since the optimizer used in main.py is SGD, and it can also be noticed that the weight_decay parameter in SGD is also used, should it not be equivalent to L2 ?

I also further read the BNupdate func but I still did not understand why it is L1 , cuz BNupdate just adds or subtracts a constant to each gama value?

Please help me out on this point, thanks a lot!

EddieEduardo commented 2 years ago

Clear