2013-07-19

Perceptra Test Run – 1 19th July 2013

I tried making binons at level 2 that were combinations of the property binons (ratios) at level 1 as though all the level 1 binons were independent. This was done by first creating pairs of level 1 binons independent of their position relative to each other (independent). Then these level 2 binons were combined provided they contained an overlapping, shared, common level 1 binon. This effectively produces all possible combinations of two, three, four etc. level 1 binons. Naturally it was limited to only create ones when the two source binons occurred coincidentally in the same stimulus and were already familiar from previous stimuli.

The result is the creation of many binons, as illustrated below. The learning rate steadily increased as the lower level binons became ambiguous. There were quite a few "no predictions"