Alejandro de la Vega

Hey, next week would be better actually, I'm moving this week. @rbroc, want to chat about this too?

Yep, agreed on both fronts. About the `Replace`, we may want to disable predictors that are categorical then, because they otherwise are going to impossible to use. We could also...

It would be interesting to try face emotion, and see if its any good though. We could easily do it through `pyNS` for now. At least we could warn users...

@jdkent: - for some notion of how many anatomical compcor components to include, looks like the first 5 or enough to explain 50% of the variance were used (in resting...

Actually, I'm noticing that a model w/ STFT is taking a very long time to compile. I think 0.1 might still be too big of a hop size. Maybe 0.5-1...

I think for now I'm going to delete all STFT predictors, as they are sampled at 10hz. I think 1-5hz would be more reasonable, and would make our db smaller...

@peerherholz when you have list of extractor you'd like me to extract, comment here @rbroc can you also take a look at Pliers, and suggest any audio based extractors you're...

Awesome, thanks for the suggestions. @rbroc lets talk about this next week maybe? peer is interested in running some of these models so we could also all meet and discuss...

Awesome. @satra do you have any suggestions for how finely we have to sample the final predictor? I'd like for it to be as coarse as possible for space saving...

Thanks. I think we can do 10hz but I should investigate a bit first. The problem is not only space but that the queries with lots of data can be...