norms of actual vs. fixed repertoire

[mobeets]

pressure towards lower firing rates would be evidenced by smaller norms of actual null activity vs. that predicted by fixed repertoire.

data is in factor activity...

so for each movement direction group, I compare the mean norm predicted vs. observed:

each point is a movement group from one of 41 sessions.

now, just taking the mean of these per session, so now each point is a session:

[mobeets]

BUT! factor activity is centered around zero spikes. so we should actually first convert predictions to spikes, and then take the norm...

below, this is converting latents to spikes under the generative model, but with no noise added.