Sunday, September 18, 2011

The Amazing Power of Concurrent Pattern Recognition

The Key to Sequence Learning

In view of what I wrote in my previous post, some of my readers may be wondering what it is exactly that I discovered. Without giving away the secret, let me say that the key to learning sequences of sensory events is to come up with the right method for concurrent pattern recognition. Once you've got that figured out, then sequence learning becomes a breeze. In fact, sequence learning becomes just a recording process in which each recorded event is a recognized pattern of concurrent signals. And as with any recording process, precise and correct timing is paramount.

That being said, this is not what makes sequence learning easy. What makes it easy is that, with the right concurrent pattern recognizer, there is no need to search the sequence space for good or valid sequences. In other words, every sequence is a good sequence. Just record them all and let the Branch mechanism sort them out.

This is all I can divulge at this time. Later.

See Also:

I Was Wrong About Pattern and Sequence Learning


Bill said...


If you are not already familiar with Binary Decision Diagrams (BDD) you should check them out:


Will S. said...

I def. want to know more about this. I'd imagine what you've implemented -is- similar to what the previous poster referenced. Is that the case?

Louis Savain said...


Thanks for that reference. I had some familiarity with binary decision trees before but never thought much about the subject.

Will S.,

The answer to your question is no. One of the things that I have come to understand about memory is that it is more than a just recognition engine. Consider that everything that we hear, feel and see, although caused by the environment, is not in the environment. It's all in the brain. This means that memory is an extremely malleable structure that can almost instantly change its parameters to accurately reflect changes in the environment. In that sense, memory is not the source of our intelligence. The logic that drives our cognition resides in the environment, not in the brain.

It occurred to me that this environmental logic is already inherent in the concurrent patterns that we detect. As a result, concurrent patterns are extremely constrained in the way that they are manifested. In other words, the number of possible concurrent patterns that can follow each other in a sequence is minuscule. Most of the times, there is only a single successor to a given predecessor. This is why I have said that, once you figure out the proper way to detect concurrent patterns, sequence learning becomes a breeze.

So the real problem in sequence learning is not determining what follows what. That’s easy. The real problem is to figure out how to build a structure that can instantly record the changing temporal intervals between concurrent patterns, as they occur, and then use these recorded intervals to make predictions about what will come afterward. Context is essential to prediction. There is no need for complex Markovian or Bayesian statistics.

Of course, there is also the problem of attention but attention is really part of recognition. And recognition is just a matter of activating a single branch in the tree of knowledge depending on the context. And, to repeat, the tree of knowledge is not a binary decision tree. It will all become clear when I publish a full document on Rebel Cortex.

However, that will have to wait a little while because I need to raise some funds. I am not exactly sure about how to proceed in this matter. I may just sell the documented source code for Rebel Speech to whomever is willing to pay a premium for the technology. Or I may form a startup company to incorporate it into various products. We'll see.