[ USCBP || FARS || up || down || prev || next ][ model || bio || impl || sim || doc ]

1. Micro-Schema


micro-schema Fig.1: Micro-schema structure.

Most of the cortical regions in the FARS model are implemented as a number of micro-schemas. Conceptually, a micro-schema (Fig.1), or referred to as a column, is an atomic computation unit composed of a set of four leaky integrator neurons:

The signal unit detects sensory input from other columns. Activation of the signal unit can causes the output unit to fire, in which case, the column is active. The activity of the column is subject to two other constraints, whether the column is primed and receiving non-negative support. Priming signals from higher-level regions are detected by the priming unit, while activation and inhibition from other columns are received by the support unit. If both constraints are satisfied, the column is prepared to fire when correct sensory input comes in.

For more details on the motivation and functional structure of micro-schemas, please refer to "Micro-schemas: An Intermediate Computational Mechansim" in Chapter 5 of Fagg's thesis.

real column Fig.2: Actual column.

In the actual coding of the model, columns are implemented a little differently (Fig.2).

A column still has the four basic functional units (priming, support, signal/sensory, output). However, the priming unit and the support unit are implemented in a structure called gate, and a column can have a number of gates. The output of the support units in these gates are collected by a collector unit. The collector unit and the sensory unit together determine whether the output unit should fire.

For more details on the column dynamics and implementation, please refer to: