Open studywolf opened 11 years ago
We don't absolutely need this for the basal ganglia. If we solve for the decoders the normal way, we get pretty much all positive or negative weights, just due to the nature of the representation. So the models run fine without this option and are very close to what we'd get with it.
When we do get around to implementing this, though, it should also have some mechanism for caching the decoders. It can't use the current mechanism (which caches A and Gamma_inv).
Cool, glad to not have to implement this for a matching basal ganglia!
Implement function that allows you to specify only positive or negative decoders, used by the basal ganglia.