Summary: weekly mtg 20170627 (Byron, Steve, Aaron, Emily, me) - mobeets/nullSpaceControl GitHub Wiki

To discuss:

  • potential reviewers
  • specific manuscript questions
    • Results first para outline
    • "For all timesteps" comments
    • IME: where does it go
    • SSS: how to handle the reaction "if it's a decrease, doesn't that support minimal energy?"
    • discussion section on changing output-null activity
    • where to cite supp figs
  • references
    • where to cite Strick
    • where to cite Vivek
    • where to cite Law & Schieber

Sentence "you use all patterns within the repertoire that will produce the behavior"

  • this should be one of our main points
  • make sure this is mentioned in the fixed repertoire hyp. section
  • maybe mention this in SSS section as well?

Discussion section:

  • changing null space: maybe cut the first point, put the second point somewhere else
  • maybe remove section headings: do things still flow?
  • add "different cuts" figure for two mappings?
  • dependence on different kinematics (like, what if we had twice as many R targets as L ones)
    • difference between support and distribution (maybe this is a whole para, after the habitual para maybe? mention how we have a symmetric-ish task, but maybe our predictions wouldn't be as good if this were not the case...)
  • habitual para: connect last sentence to the one before it; last sentence not clear how it relates
  • is it clear that we see Cloud as being similar to unconstrained?
    • in the OFC section, after we list the differences, maybe list the way that we're similar to uncontrolled, in that once you include yoking and the repertoire
    • try the same set-up for habitual: here's how we're different, here's how we're similar
  • these points should just stay in the same section, but rephrase that the section is more about caveats, ways that our results might be different in this case, blah blah; e.g., this study focuses on short timescales; it would be interesting to see if other principles came into play in long-term timescales; and the other point is that we have a symmetric task with equal numbers of trials to different target
  • also still remove section headings

Results outline:

  1. First section
    • "We used a BCI with 85-94 units from three macaques..." etc. etc.
    • Differences between cartoons and actual # of dims and dirs
    • FA
    • IME?
  2. Minimal energy
    • ...

Results:

  • Goal: reduce to 1/2 to 2/3 size
  • e.g., the 88 part is unnecessary since our first step is to do FA, then we just get 8
  • keep IME here
  • "For this visualization..." --> move to caption, and also bring up for Unc hyp
  • little mini sentence at end of results to refer to S2, S3, S4

To do:

  • SSS: why it's not agreeing with minimal energy
    • cloud sig. diff. from data by monkey?
    • maybe mention magnitude
    • maybe soften the < 0, > 0 comments to allow for "substantial" decrease, "substantial" increase, that way the magnitude being near zero seems relevant
  • ref Strick in the discussion section on why minimal energy might not be about firing rates

Third point of paper (end of abstract, intro, etc.):

  • "The dominant constraint on output-null activity is the repertoire. But from within the repertoire, subjects will choose any activity consistent with desired behavior."
  • this should be mentioned in the cloud's hyp section, so we're connecting where we see this point in results
⚠️ **GitHub.com Fallback** ⚠️