Hi Nupic, I understand that there is something equivalent to an encoder that transforms data coming in through the body's senses into sparse representations that get passed on up in the brain. But I am unclear about how each layer creates its own SDR to pass up to the layer above it. For example, let's say that the SDRs for the letters in the word "the" are the following:
t = 1110 h = 1011 e=1100 So the feed forward input to the next layer above this layer to create the SDR for the word "the" would be: 111010111100 What mechanism would convert this into an SDR in a CLA with hierarchies?(another encoder?) And what prevents the abstract higher layer SDRs from getting so sparse that they become meaningless? Mika
_______________________________________________ nupic mailing list [email protected] http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
