2005-04-28

A better day

Today is a better day.

I have had doubts about the simple approach which ignores the six layers of the cortex.

I just read two papers written by Dileep George and Jeff Hawkins: A Hierarchical Bayesian Model of Invariant Pattern Recognition in the Visual Cortex. and Invariant Pattern Recognition Using Bayesian Inference on Hierarchical Sequences

And while there is a mention of a match between the Bayesian network mechanics and the biological happenings in the six layers, the experiments described do not replicate directly what happens in the brain at that level.

I might have to revisit the modified Markov models proposed in my degree paper however.

As soon as I get in a phase where trials can be made using prediction, I might decide that the signals traveling up and down are not simple values, but a distribution of probability of several values.

It's interesting to see how things converge:
  • the hierarchical concept-forming idea of mine (maybe someday I post them here) seems to resemble very closely the hierarchical structure of the memory-prediction framework,
  • the statistical NLP knowledge I accumulated and used in the degree paper seems to be of central importance in the learning and prediction of the memory-prediction framework.

0 Comments:

Post a Comment

Links to this post:

Create a Link

<< Home