- Two rather surprising results from the lesioning analysis of the model: 1) Removing the behavioral modulation module led to only a modest (2.8%) reduction in predictive accuracy. Surprising given the known effects of locomotion etc. on mouse visual responses. (1/2)
- I suspect that behaviour seems unimportant because normalised correlation averages over repeats, minimising the effect of trial-to-trial variability. Single trial correlation should show a bigger difference, we observed something similar in: openreview.net/pdf?id=qHZs2... (Table 1 vs Table A.7)Apr 10, 2025 20:02
