search

emer / axon

Axon is a spiking, biologically-based neural model driven by predictive error-driven learning, for systems-level models of the brain
BSD 3-Clause "New" or "Revised" License
20 stars 7 forks source link

BGThal should have a modulatory effect on PFC #202

Closed rcoreilly closed 1 year ago

rcoreilly commented 1 year ago

Biologically, BG thalamic projections are likely "matrix" like, targeting distal dendrites in layer 1, where they hit both excitatory and inhibitory targets, with a net result that is probably best approximated as multiplicative or modulatory.

Including this in the pcore / boa model would improve robustness of gating, enabling activity to be strong when thalamic activity has a positive multiplicative effect, but cutting it off otherwise. anatomically, distal dendrites will be for superficial and deep layers, affecting other inputs in these areas. critically, it may be that the NMDA maint projections are relative unaffected, enabling sustained maintenance to continue once the phasic gating input goes back down. We likely need a different marker to separate the other inputs from the recurrent maintenance connections (MaintG). not a bad idea to have those separately for driving stronger NMDA separate from other inputs in any case.

Also fixed an issue with SC, where if CSGated, and we go to the next trial to drive the action, ACh is still active (insufficient adaptation in SC) and it activates BLA extinction mechanisms!

rcoreilly commented 1 year ago

MaintG is now implemented and now BGThal has full control over PT maintenance, and we can even have strong Super -> PT projections which was impossible before, because it would just drive MaintEarly PT activity without BG gating control.

This also nicely handles the problem with novelty persisting -- not happening at all now.