Summary: weekly mtg 20170607 (Byron, Steve, Matt, Aaron, Emily, me) - mobeets/nullSpaceControl GitHub Wiki

• minimal-energy implementation
• habitual -> "minimal intervention"
• discussion outline
• predicting tuning changes

To do:

• use factor version of minimal firing, because the spike version is intractable, and what's it matter anyway
• IME in main text
• flesh out discussion
• send manuscript to all authors! (expect 10-14 days)
  • We're about to submit this to Neuron. Please take a read. Everything is finely tuned except for the discussion...We'd greatly appreciate it if you could get back to us by [10 days from now]

Tuning changes:

• tuning in perturbation using cursor movement direction: more circular, because you're asking what the activity would be when you know it has to move it in a particular direction

• reassociation assumption: movement-conditioned covariance is the same as it was before

• steve: empirical tuning changes were remarkably small compared to what reassociation was predicting; i.e., how is it possible to have tuning changes that are this subtle but to have this big of a behavioral improvement

• maybe propose a framework for what we're trying to do?

  • e.g., factors first, p.d.s first
  • what info we have access to
  • how we're assessing success, errors in prediction, etc.

• simulating errors, relabeling points

• ask steve how it was that the washout effect could be not consistent with the null space stuff