2015-11-24

Perceptra 1D Test Run – 1 - Date:  24th Nov 2015

I have put the binon tree activation and stem binon code from Perception which is processing Morse code sequentially into the Perceptra 1D which normally processes the pixels of handwritten digits in parallel. The results of the parallel processing of the pixels in the 1st handwritten digits (a “0”) are shown below. It generates 18 contrast binons of which the last one is at level 4. The result of the sequential processing however starts acting differently when processing pixel number 17.  After pixel 16 the attended to binon (the one at the end of the perception list) is the level 2 combination binon #6. This results in the stem binon being cleared because of the difference in levels between the trigger in the stem binon and this binon #6 being attended to. So at position #17 the previous stem binon level 1 trigger is no longer available to form the combination level 2 binon #8 that was produced in the parallel process.  The parallel process passes over all the pixels at level 1 and then all these results at level 2 etc. The sequential process builds higher level binons as it proceeds sequentially through the pixels – only one pass.

Can I change the sequential process so it produces the same results as the parallel one?  I could try not clearing the stem binon information if the attended to stimulus is not familiar. In other words if the attended to binon is at a different level only clear the stem binon information if it is also familiar.