Skip to main content

[Pilot] Multisession take 2

Created on March 1|Last edited on March 2

Sorted data, 96 channels, hold eval set constant.

No obvious scaling.

123456789102030405060708090100200epoch0.340.360.380.40.420.44
Run set
3



Unsorted, constant eval set.

  • We do still have scaling per expectation - the value is just quite small.

Run set
5

Restricted test session trials to 256.
  • Sees definitive improvement over single session, and definitive lagging in permuted.

Run set
8