Summary: weekly mtg 20170711 (Byron, Steve, Aaron, Matt, me) - mobeets/nullSpaceControl GitHub Wiki

Tiny questions:

  • Holm-Bonferroni: "how many comparisons"?
    • remove multiple comparisons correction
  • Mention direction vs. speed decision in results?
    • maybe reference experimental procedures
  • Methods: mention Byron's FA code?
    • don't bother
  • Background: prepare code pack with example code
    • not now; in the future
  • Capitalize names of hyps? "uncontrolled-uniform"
    • aaron doesn't care anymore
  • "Observed vs. empirical" - confusing in "unc-empirical" section
    • stick with observed
  • Methods: how much detail to explain WMP selection?
    • stick with just "permutation of factors" and "updated value of B"
  • Future to do: how this work improves on what Matt's paper shows
    • i.e., it shows the entire repertoire is used
    • question: how many points of int. repertoire does cloud not sample

Bigger discussions:

  • Aaron's SSS suggestions
    • focus on reframing this as more an extra test of Cloud, not just as a control
  • Title: Strategies for resolving neural redundancy
    • Towards resolving neural redundancy [maybe too weak]
    • Constraints on neural redundancy
    • some way of starting with "Neural redundancy..."
  • References:
    • Athalye: add where we mention FA, maybe also Churchland (Nature Neuro, 2010), Harvey, Tank (Nature, 2012; Choice-specific...)
    • Law & Schieber (near where we first mention Kaufman, with null and potent); also some Shenoy paper with "null space" in the title (Stavisky, Neuron) should go in the discussion para
    • Mussa-Ivaldi
      • overlapping dimensions paper: goes in para on function of studying null space activity
      • learning to be lazy paper: maybe about the minimal energy section
  • histogram error metric not invariant to rotations

To do:

  • maybe point how not using IME doesn't make predictions that much worse; still, predictions are better when we condition on the potent dims, as opposed to just any two dims
  • SSS: in reverse for empirical data only, say
  • poster Aaron mentioned about repertoire, arm control, something, yoking??