Summary: weekly mtg 20161122 (Byron, Steve, Matt, me) - mobeets/nullSpaceControl GitHub Wiki

manuscript

  • how to quantify muscle-inspired hypothesis failure in text?
  • figure 6: mean and covariance error for each of two (three?) monkeys?
  • where to fit in empirical versions of baseline/minimal if no supp. text? just as an extra figure with them included?

Figure 6:

  • think about color groupings: e.g., MF and BF, UC and UE, Hab and Cld?
  • acronyms in names?

When discussing the two mappings hyps

  • say as an aside, we also considered empirical versions of MF and BF, and these didn't do as well (see Supp)

Best mean hypothesis

  • after showing MF and BF results, but before moving on to the two mappings
  • in Results text: "You might be worried about this..."

Scoring muscle-inspired hyps:

  • something that's just about marginal histograms
  • something as intuitive to understand as possible
  • like, percent overlap of each histogram on average
    • something that ranges from 0 to 1
    • absolute difference between the two histograms, divide by 2
      • so that 0 is non-overlapping, 1 is perfectly overlapping
  • basically, do they match or not

nature neuro vs. neuron supp. figs

  • nature neuro: ≤ 10

  • neuron: ≤ # main text figs+tables

  • no supp. text

  • check for recent issues to decide on this

NDSEG essay feedback (500 words)

  • ask Ben and Will for NDSEG essays

behavior as patrick's instead of cursor progress for sure?

  • Steve okay with progress, because we want mean improvement, and less variance
  • in main text, it's just "we selected a subset of trials where we saw learning with asymptotic behavior"
  • have the figure 4a, and introduce both the motivation for IME (getting error all the way down to intuitive levels)
    • don't have to talk about or use cursor progress at all here; that can just be our session selection criteria?

go for formatting for Neuron

send email with manuscript