Summary: weekly mtg 20160803 (Byron, Steve, Matt, me) - mobeets/nullSpaceControl GitHub Wiki

Discussed

eliminate cursor magnitudes near 0

flow diagram for each hypothesis

1. take a timestep of neural activity

Fig. 2

• uncontrolled after minimal/baseline
• but looks good otherwise

overloading of color in Fig. 2

• choose one color, use discrete gradients of color
• e.g., all green

results fig

• use 16 thetas
• alternate presentation ideas:
  • for single target, column for all dims, real data only
  • or for single dim, all target angles
  • or ticks of means in 16x8 grid for all hyps
• smooth MORE, no jags; need to fit all hyps on same plot
  • true data is thick
  • and don't normalize the heights; let it go off the y-axis? or chop axis itself

results fig: mean/cov bars

• this doesn't convey the fact that NONE are good
• maybe show hab/cloud anyway? and say we'll introduce them soon?
• or just report mean errors per session
  • with bars in supplemental
• and then later, show all

results fig:

• scratch bars from main figs
• add to supplemental bars with ALL hyps
• show scatter of cloud vs. hab in mean, and then another for cov
  • or scatter of mean, cov as x-y axes, and show all hyps; z-score mean and covs
• this frees up results in terms of histograms

steve's special space

• check without IME? trend with learning should get stronger
• compare overlap in row spaces--explains trend with learning?
• SVD on the 2d special space to get orthogonalized basis
  • i.e., take the SVD of the 10x2 mapping: (NB*NB')*RB

To discuss

• learning trend
• irrelevant/relevant
• valid spiking

---

learning trend

CLOUD:

HABITUAL:

irrelevant/relevant

blue is jeffy, red is lincoln

valid spiking

every latent in hypotheses is valid if this means that there exists a spike within min and max spike count that would produce that latent. so instead say a latent is valid if the mean spikes from the generative model are within 0 spikes and ~40 spikes.

for each date (rows), here's the percentage is says are invalid:

true     hab   cld   unc   base   min*
    12    12     6    22     4   100
    13    17    16    17     0   100
     0     2     1     2     0    100
     1     3     2     5     0   81
     8    11     8    12     0   72

If instead I use empirical min and max bounds from intuitive block...

true     hab   cld   unc   base   min*
     3    12     3    25     0   100
     4    11     4    18     0   100
     1     6     2    12     0   100
     9    20     6    28     0   100
     5    20     5    33     0   100