Open gretatuckute opened 2 years ago
is this related to our brief thought about say, paragraph semantics, stimulus features etc affecting brainscore? how do we do this in an automated way?
initial motivation was due to the bigger question of "why are certain models good and how do we easily quantify that" but it is true that one needs to come up with a stimulus set that is somewhat representative so quantify layer / model properties
given an ANN / layer, quantify those activations --> understand WHY certain models/layers are good