[Korrelan]

Hi everyone, just thought I'd drop in and say hello.

I'm still busy working on my project & making progress slowly. 

[Korrelan]

Hippo - Korrtecx learning (mem) phase.

Input = 180 sets X 5K patterns (P)

Graph shows ALL sparse outputs, single spike shows confidence/ accurate recognition (R) of current P.

Pass
#1 R of 180
#2 new P set, no R
#3 learn new P, notice map colours change
#4 R of new P

[MagnusWootton]

so your learning a data set into it?

I know how to do that too,  but I didnt actually get it working in the robot yet,   but I got a low error report, but I didnt use it for anything.

It said I managed to fit 5000 records in 10,000 synapses into a single braincell/perceptron,   at about 10% error.  which would be good enough to run a step function with,   but I never managed to use it,   so I'm going to check it out some more If I can get some work done.

[Korrelan]

Noise - there is no 'stray' noise, the mechanisms behind thalamocortical grounded 'holistic' memory requires/ generates massive levels of transient signals (TS) within all the cortices.

Single peak shows accurate recognition in spite of 50% TS bursts.

[Korrelan]

Holistic - integration of patterns, though this is not a normal pattern, this forced visualisation shows how unchecked (no inhibition) neocortex lateral connectivity/ propagation links cortices.

Any singular voxel (1K+ neurons) can integrate into the whole.