Taesu Kim
Results
1
comments of
Taesu Kim
Our implementation is basically not much different. The original one only considered output layer as trainable parameters to calculate influence and it only applied to their experimental networks, but ours...