search

interpreting-rl-behavior / interpreting-rl-behavior.github.io

Code for the site https://interpreting-rl-behavior.github.io/
Creative Commons Attribution 4.0 International
0 stars 0 forks source link

Update xcaus plotting code so that we can create plots for sets of samples that have been filtered according to hx_activations #72

Closed leesharkey closed 2 years ago

leesharkey commented 2 years ago

The goal right now is to find IC to IC stories that we can confirm with gradients. We find the correlations using xcorr plots but confirm/disconfirm them using filtered xcaus plots.

Why filtered? The idea here is that it should group samples by a feature that we want to causally study. For instance, if IC6 is highly active, then on those samples we'd like to see what has caused it to be so high/low. But we'd also like to group the plots for when IC-K is high/med/low at t=2 in order to see what effects IC-K being high has in later timesteps.

leesharkey commented 2 years ago

Done in 89be525753b5366ddea8913c2e4c79ace7f32d7d