transfer learning on u

[georgemilosh]

I think there is no reason to have transfer learning work on u, which seems to be the case provided that load_from='last' which is default. I guess I could turn of transfer learning completely by using load_from=None. Maybe that should be defaault behavior. Or maybe default behavior should be that transfer learning occurs only with tau. Otherwise we can easily fall into a trap when not performing analysis of hyperparameter optimization carefully. I am a bit about worried about this transfer learning.

[AlessandroLovo]

When you do hypop you should disable transfer learning by setting load_from=None.
In principle you can transfer learn from all runs that have a compatible network architecture.
I guess the only sensible thing to do is to put load_from=None as default.

[georgemilosh]

Yes, unless you are hypop while transfer learning from tau.

[AlessandroLovo]

In that case you can pass load_from='last--u__same, as explained in the docstring of Learn2_new.get_run. This will transfer learn from the last run that has the same value of u.