Summary: weekly mtg 20161221 (Byron, Steve, Matt, me) - mobeets/nullSpaceControl GitHub Wiki

To discuss:

  • What purpose does Figure 1A serve? Could we maybe combine Figures 1 and 2?
  • How can I keep the figure captions shorter? Or, we could break up Figure 3 (and 5) into cartoon fig and histogram fig?
  • Should Supp. Fig 1 be a main text figure?
  • How can we introduce IME prior to presenting results?

Discussed:

PNC affiliation

Highlights:

  1. ...choosing between behaviorally equivalent neural activity "Multiple patterns of neural activity lead to the same motor act."

Figure captions: not a walk-through; just an explanation

  • e.g., Fig. 3A

Fig. 3E: xlims don't match the others

IME: key is that the subject knows these are behaviorally equivalent patterns, so we use IME

  • okay to introduce IME afterwards

Histogram overlap %: make it into an error, not overlap

  • break it up by monkey as well
  • add a top row to Figure 6 for histogram error
    • the intuitive hyps can still be sup
  • for intuitive hyps, change it back to "none performed better than _%"
  • and in supplement, make sure to mention how these are different from those presented in Figure 6

Don't include Minimum-sample and Baseline-sample at all

Baseline firing: does well in histogram overlap because it counts all dimensions equally

  • maybe make (supplemental?) figure of overlap % without top dimensions
  • could say "going to focus on the top 95%"
  • and error vs. dimensions (cumulative: error using first dim, first two dims, etc.)
  • error weighted by total variance in that dimension

Figure S2

  • Fit learning to angular error, do things change
    • if so, let's just use that, for easier comparison to IME
  • Also, for Figure S2, maybe also add in percent correct for the same session
  • just use corrects for fitting

Figure S4

  • might break into two figs, for the two different types of runs
  • also include histogram errors

Figure 2

  • ignore A, v_{t-1}, c
  • just match Bz_t
  • basically just remove all equations
  • in text: "key concept we want to convey is that there are multiple z_ts that have the same Bz_t", and so there are multiple neural activity patterns that result in the same movement
  • change z_t font
  • basically, no reason to have equations in main text