Closed rcoreilly closed 1 year ago
this is also evident in the pvlv posacq case -- takes a long time to kick in and is noisy.
The use of GeIntMax for recv learning activation was problematic in multiple ways (too similar across neurons, underlying GeInt wasn't being reset at start of trial so there was carryover). Now using CaSpkD if above threshold, then falling back on GeIntNorm as normalized GeInt with clean integration. The long tau of GeInt allows it to do most of what Max was supposed to do, and also has more differentiation and sensitivity to sustained activity.
See #306 for discussion -- now working well.
really bounces around a lot. and never gets up to match Rew levels.. needs more work.