Fig 2. General HPC architecture (adopted from Rao. [14]).

a) On arrival of input, predictive estimator module at each higher visual processing level makes the estimate and sends prediction signal to its next lower level by feedback connection and receives the corresponding prediction error by a feed-forward connection. The error signal is used by the predictive estimator module to correct the estimate for better prediction. b) General predictive estimator (PE) module constitutes of (i) neurons to represent the estimate of the input I by their response vector r by minimizing the bottom-up (I − Ur) and top-down (r − r^td) error, (ii) feed-forward error carrying neurons has the efficacy matrix U, which encode the basis vectors their synaptic weights (or receptive fields), (iii) prediction Ur carrying neurons and (iv) top down error detecting neurons.