jakesnell / prototypical-networks

Code for the NeurIPS 2017 Paper "Prototypical Networks for Few-shot Learning"
MIT License
1.12k stars 254 forks source link

Train accuracy is lower than val accuracy #16

Closed Paul0629 closed 5 years ago

Paul0629 commented 5 years ago

I've tried two versions of prototypical-networks source code. Both of them get me this result, train loss is bigger and train accuracy is lower.

zxkplus commented 3 years ago

yes , I try to reproduce this code , but I have about few percentage points grap in 20 way 1 shot , it is about 0.93%。how are about you?