Friday, August 24, 2018

Temporal Correlations and Predictions in the Cortex

Using Minicolumns for Predictions and Temporal Correlations

This is another short post. I have written previously about cortical columns. I claimed that the minicolumns that comprise cortical columns are highly predictive event detectors. That is to say, the firing of a minicolumn is predictive of the firing of a minicolumn residing in another cortical column. Minicolumns owe their predictive abilities to the fact that they detect rare temporal correlations.
Note to believers: The Book of Zechariah (Zech 3:8) uses the Hebrew word 'mo.phet' which means wonder, miracle, symbol or sign (of a future event) to describe the activation of a 7-node minicolumn. That is to say, the activation of a minicolumn (the stone with seven eyes) is a rare event that announces another event.
Temporal correlations run the gamut from highly correlated to occasionally correlated concurrent signals. The signals that arrive at the entry or bottom layer of the neocortex come from highly correlated patterns. As one climbs up the cortical hierarchy, the minicolumns detect less and less frequent correlations. It just so happens that the less frequent a correlation is, the more deterministic it also is. A fully deterministic prediction is one that has only one predictor. In other words, minicolumns at the higher levels of the cortical hierarchy are very deterministic.
A cortical column shown with five minicolumns
This is all for now.

See Also:

Fast Cortical Learning Using Spike Timing


Louis Savain said...

Addendum (for believers only):

I wrote in the past that a minicolumn has seven inputs, one of which is the main or primary input to the parent cortical column. The primary input is common to all the minicolumns that comprise the parent column. The primary input of a cortical column is symbolized in the book of Zechariah by the high priest Joshua. The other inputs are his associates (also priests) that serve with him in the house of Yahweh.

The bottom layer of the cortical hierarchy receives inputs from pattern detectors in the pattern layer. The question is, which of the inputs should serve as the primary inputs to the cortical columns? Zechariah gives us a powerful hint. He describes Joshua as a "brand plucked from the fire." What this means is that Joshua outshines the other inputs. That is to say, the primary input is much more active than the others.

If you are a believer who's trying to decipher the occult texts, it's important to keep this in mind when reading the vision of Zechariah.

Louis Savain said...

Again, I cannot emphasize enough the importance of partial temporal correlations. Pattern memory learns highly correlated elementary patterns. These are patterns of signals that occur together frequently. However, there are signals that are only occasionally correlated. For example, signal A may be correlated with signal B only once every few spikes. Even though they are less frequent, partial correlations are still very important to be able to detect. Without this ability, we would have a hard time understanding music, speech or the 3D structure of the world around us.

Normal pattern detectors cannot detect partially correlated signals. This is the job of the cortical columns and their minicolumns. The fact that the human cortex contains over 100 million cortical columns is a testament to the importance of partial correlations.

This is my current understanding of the cortex. There are aspects of it that I cannot divulge at this time. And there are still some things about it that I do not understand although I have no doubt that I eventually will. I've come a long way.

Louis Savain said...

I'm working on a new article on pattern learning in the thalamus that does not use a hierarchy. I was wrong about the need for a hierarchy in the thalamus. It happens.

Thalamic pattern learning is the discovery of highly correlated elementary patterns. As I explained above, it's important to distinguish between correlated and partially correlated patterns. Coming soon.

Louis Savain said...

I have an article on pattern learning ready to post but I'm having second thoughts. I think I'm moving too fast. I fear that someone else (a bad person) may gain enough understanding to extrapolate the whole kit and kaboodle, so to speak. Maybe it's time for me to change gears and start writing about occult physics again. Decisions, decisions.