This was semi-fixed by @colah and me a while ago. We decided against parameterized noise comparisons and just focussed on the practical benefits of having transparency in your neuron visualizations. I will close this issue for now, but I am open to having a broader discussion on whether that's the "right" story to tell here.
This was semi-fixed by @colah and me a while ago. We decided against parameterized noise comparisons and just focussed on the practical benefits of having transparency in your neuron visualizations. I will close this issue for now, but I am open to having a broader discussion on whether that's the "right" story to tell here.