Prep: weekly mtg 20160323 (Byron, Steve, Matt, me) - mobeets/nullSpaceControl GitHub Wiki
bug for within-trial alignment: bug for volitional, minimum, baseline (but not serious)
simulations:
- latents to spikes to latents:
U = Lz + diag(ph)/L'
instead ofU = Lz
- roughly, it seems that adding noise to latents make the covErrorShape better, but the covErrorOrient worse.
- and curiously, the volitional errMean seems to be the same in hab or cloud sims, but the covError is somehow basically dead on.
- also, for cloud-sim data, the covErrors for habitual seem to be the same as the true data; not so for hab-sim
ime
- outperforming cursor errors in intuitive block
- mild hypothesis improvement?
- also, volitional hypothesis totally blows up for some reason (like, from errOfMeans of 6, to 1000)