search

liuzechun / Bi-Real-net

Bi-Real Net: Enhancing the Performance of 1-bit CNNs With Improved Representational Capability and Advanced Training Algorithm. In ECCV 2018 and IJCV
179 stars 39 forks source link

请问你们提出的“Magnitude-aware gradient with respect to weights”和xnor中有什么区别？ #13

Open pppLang opened 5 years ago

pppLang commented 5 years ago

看起来sign的系数好像都是weight的一范数的均值

flymmmfly commented 1 year ago

感觉没啥区别