The research article http://www.dartmouth.edu/~rhg/pubs/RHGtc4s.pdf says,
Table 2. Simplified steps in thalamocortical operation. i) First input from periphery topographically activates core nucleus (Ct). ii) Ct topographically activates corresponding middle layers. iii) Activated middle layer modules vertically activate corresponding suprajacent layers. iv) Output from superficial layers topographically activates deep layers. v) Diffuse feedback from L.V output to matrix nucleus (Mt). *vi) Topographic feedback from L.VI output to NRt and to Ct, selectively inhibiting the portion of the* *input corresponding to the cluster response.* vii) Next input (or portion of input) arrives; Ct topographically activates middle, superficial, deep layers. viii) Layer V receives combination of nontopographic input from Mt, produced by prior input, and topographic activation from superficial layers produced by current input. ix) Intersection of these inputs selects sparse L.V response and synaptic potentiation. x) Repeat steps v) to ix) until input completed. >From this article's conclusion, previously learned sequences controls the current input. The inhibition time is longer than excitation time, the columns activated in the previous steps inhibiting the current input for that column. In this way, even if same static image shown to the brain, the brain clustering that input and forming the sequences of input. I am seeing this one as type of encoding process brain does it for inputs. Correct me if i am wrong, Thanks Ramesh G.
_______________________________________________ nupic mailing list [email protected] http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
