Pattern completion through thalamo-cortical interaction

The thalamus is the main relay station for sensory information to the cortex. However, it receives ten times more information back from the cortex than from the senses. We have implemented a neural network that mimics thalamus and cortex connections. This architecture can elucidate the meaning of corticothalamic backward projections in the context of a reverberating dynamic leading to pattern completion and consisting in the repetition of four processes: separation of pattern components in the thalamus, linkage of these components and template competition in cortex, and feedback of the winning template to the thalamus is added to the initial sensory pattern.