best-mean predictions

[mobeets]

the latents picked by best-mean, for each session:

(distance from 0 is distance from baseline rate in latent space)

as you can tell from the above, the best-mean is close to the baseline rate. this is true whether you look in spike or latent space

[mobeets]

i tried making a hypothesis that finds the best mean per target, and it does do a little better than best-mean--about as good as uncontrolled-empirical.