Explanation AI notes Reference:https://burakhimmetoglu.com/2017/08/22/time-series-classification-with-tensorflow/ Reference paper: Explaining nonlinear classification decisions with deep taylor decomposition
I add PatternNet and PatternAtribution code. But, the patternNet result is very bad, while PatternAtrribution result is vary good. I only realize the fully connected network with Relu activation. I'm now finding the reseaon why PatternNet is so bad.
If you have any idea, please feel free to contact with me. We can share idears together.