Summary: weekly mtg 20170118 (Byron, Steve, Matt, me) - mobeets/nullSpaceControl GitHub Wiki

emphasize factor space as being nothing special (in text)

• demystify it
• because otherwise, projecting to 88d is still going to be the same 10d space
• okay, so make it clear that even doing this analysis wouldn't make sense
• could also look for indirect neurons in recordings, i.e., neurons with low loadings
• he was also proposing a different experiment, with indirect neurons
• aim for two very clear and concise experiments up front for why reader shouldn't worry:
  1. when you do FA we chose factors (orthnormalized) so that the units are the same as in the original data -- it's just a weighted combination of neurons such that we have spike counts per unit bin
  2. in the experiment itself the cursor is moved based on the factors, so it's matched in that way
  3. we want to focus on the dimensions that matter: FA is a denoising step; focusing on the big effects
• could also show min-energy in terms of spikes as well, just to show that this is a bad way to look at it too

predicting adjacent sessions: seems like a separate scientific question; save it for later