by Charles R. Legendy, Columbia University.
The complexity of the brain, and in particular the visual cortex, is so overwhelming that, in the last decades, progress toward circuit-based understanding of brain design has all but come to a halt. Part of the problem is that the usual theoretical methods, including computer simulation and mathematical analysis, are not very well suited to the specal environment presented by the brain -- if we are ever to understand how the brain works, new methods are needed. -- --
The book on which this video is based, CIRCUITS IN THE BRAIN (Springer, 2009), describes a set of methods which seem to work quite well. -- --
The book concentrates on vision, and is formulated as a model shape processing utilizing the long horizontal fibers in layer 2/3 of the primary visual cortex. If its premises prove to be correct, it means that vision, far from being a step of passive mapping as described in older models, involves functional features usually associated with higher brain function such as a rudimentary form of learning, and the transmission of rudimentary sentences complete with syntax. -- --
The new methods amount to applying engineering principles to the creation of neuronal network designs for given processing tasks, and going to enough concrete detail to permit comparisons with experimental data. -- --
The methods include analyzing the problems of "logistics" whereby fragments of information are packaged into synchronized multi-neuronal volleys of spikes, and made to reach the localities where they are needed. The video shows the methods in action, as applied to the elusive problem of visual shape processing. It shows how the brain can combine many two-element relations between details of objects into a single many-element relation -- which is how a whole complex visual image can be seen all at once, as a unified Gestalt.