search

chr5tphr / zennit

Zennit is a high-level framework in Python using PyTorch for explaining/exploring neural networks using attribution methods like LRP.
Other
190 stars 32 forks source link

Feature request: Support for max-pooling layers #77

Closed rodrigobdz closed 2 years ago

rodrigobdz commented 2 years ago

In the code there seems to be no mention of max-pooling layers, is LRP for this type of layers already (implicitly) supported?

chr5tphr commented 2 years ago

Hey Rodrigo, Max-Pooling is supported, as its gradient distributes the relevance in the correct way for LRP, ie. winner-takes-all.

rodrigobdz commented 2 years ago

Hi, Christopher, thanks for your prompt reply. I see, proceeding to close this issue then.