Binding of drive and CS can be off

[rcoreilly]

Early in learning, the first curiosity / novelty VSMtxGo pool can drive gating even with no BLA novelty pool activity, because it is disinhibited by the curiosity drive and it gets enough excitation from the higher-level OFC, ACC reps.

Turning down the curiosity drive helps significantly, but might there be an even more robust solution?

The general issue is that high-level non-drive-specific inputs in OFCval and ACC* activate everything.. could turn them down, but they are already set to be Rel = .1.

Here's the prjn scales:

Layer: VsMtxGo
	        VsGPeTA		Abs:	  2.00	Rel:	  1.00	GScale:	  0.04	Rel:  0.50
	        VsGPeIn		Abs:	  0.50	Rel:	  1.00	GScale:	  0.01	Rel:  0.50
	          USpos		Abs:	  2.00	Rel:	  0.20	GScale:	  0.08	Rel:  0.12
	    BLAPosAcqD1		Abs:	  3.00	Rel:	  1.00	GScale:	  0.59	Rel:  0.59
							    Abs Params: .BLAAcqToGo: 3
							    Rel Params: .BLAAcqToGo: 1
	         Drives		Abs:	  1.00	Rel:	  1.00	GScale:	  1.00	Rel:  1.00
							    Abs Params: .DrivesToMtx: 1
	          OFCus		Abs:	  1.00	Rel:	  0.10	GScale:	  0.02	Rel:  0.06
							    Rel Params: .PFCToVSMtx: 0.1
	         OFCval		Abs:	  1.00	Rel:	  0.10	GScale:	  0.01	Rel:  0.06
							    Rel Params: .PFCToVSMtx: 0.1
	        Urgency		Abs:	  0.00	Rel:	  0.10	GScale:	  0.00	Rel:  0.06
							    Abs Params: #UrgencyToVsMtxGo: 0
	        ACCcost		Abs:	  1.00	Rel:	  0.10	GScale:	  0.01	Rel:  0.06
							    Rel Params: .PFCToVSMtx: 0.1
	        ACCutil		Abs:	  1.00	Rel:	  0.10	GScale:	  0.01	Rel:  0.06
							    Rel Params: .PFCToVSMtx: 0.1