motor close to done

Author: rcoreilly

citedrefs.json

@@ -0,0 +1,13 @@
++++
+Categories = ["Neuroscience", "Computation"]
+bibfile = "ccnlab.json"
++++
+
+The brain relies on **opponent** processing across many different brain structures, in many different forms, because it has significant computational advantages, including shaping learning, preserving dynamic range, and enabling modulatory interactions between brain systems.
+
+* [[Neocortex]] depends entirely on pooled, surround [[inhibition]] dynamics, which create competition and thus [[attention]] and other effects. As a result, any two representations in a given area are effectively competing with each other for the ability to influence other areas, providing a "survival of the fittest" selection pressure that is a critical element of [[constraint satisfaction]] and the broader [[search]] process of finding good ways of representing the current situation.
+
+* [[Basal ganglia]] has well-documented _direct_ vs. _indirect_ pathways that compete with each other to control the output of this brain system. As shown in the [[BG ventral simulation]], having these separate, competing pathways allows the model to weigh values for and against (pro vs. con) across a wide range of absolute magnitudes, while remaining appropriately sensitive to the _relative_ differences, which enables the limited dynamic range of neural spiking signals to go much further ([[@CollinsFrank14]]). 
+
+* In [[motor]] control, muscles are organized into opponent pathways because they exert force only through contraction. The internal neural pathways mirror this opponent organization to provide dynamic gain modulation by applying excitation to both sides of the opponent pathway, which is one important way that the [[cerebellum]] influences motor behavior. This enables a kind of _multiplicative_, modulatory interaction between brain systems, which can be much more efficient in allowing higher-level systems to exert simpler, lower-dimensional forms of control over lower-level systems.
+