Friday, December 29, 2017

A Couple of Questions About Pattern Recognition in the Brain to Ponder in 2018

Instant Recognition of Complex Patterns in the Brain

In a previous article on pattern recognition in the brain, I wrote the following:
Unlike neural networks, the brain's pattern memory does not learn to detect very complex patterns, such as a face, a car, an animal or a tree. Strangely enough, in the brain, the detection of complex objects is not the job of pattern memory but of sequence memory. Pattern memory only learns to detect small elementary patterns (e.g., lines, dots and edges) which are the building blocks of all objects. The brain's sequence memory combines or pools many small pattern signals together in order to instantly detect complex objects, even objects that it has never encountered before.
This begs a couple of questions. First, if pattern memory can only detect small elementary patterns, how is it that we can see very complex patterns? Second, how can the brain instantly see and understand a new object or pattern that it has never seen before? Obviously, there is more to sequence memory than just sequences of elementary patterns. As I said elsewhere, sequence memory is one of the most beautiful and ingenious systems in the world.

Hang in There

Have faith and all will be revealed in time. When will it happen? Answer: I do not know but it will certainly happen in our lifetimes.

See Also:

Fast Unsupervised Pattern Learning Using Spike Timing


Peter ( said...

After reading Dreyfus I'm left with quite some questions to ponder in 2018.

Lets say you detect a book. What now? Books as objects have wildy, infinite, different types of uses depending on the context. When in fight, a book is a weapon, when studying it is a teacher. (Also, is the sticker your hypothetical nephew put on your book, part of the book or is it an independent object? It depends.)

Choosing which context is relevant is a problem that regresses infinitely or it seems that way to me at least. It is arguably more pressing than just identifying the book.

Another problem is that of abstract thought and analogies. There are concepts, like straight lines or "straightness", that can be applied in general, on visual data and on sound for example. When I draw a line you can imagine a monotonous sound, when I draw a wave you can imagine a "wavy" sound. These types of thought go way beyond just these obvious examples. A person can live a straight life or you can be straight with someone. These types of "straightness" are not just wordplay, we extracted some underlying principle and apply it to all sorts of situations. Allowing us to say: "this is like that" in some vague, but useful manner. Human life is rife with these kind of analogies.

Abstract thoughts like "law", "interconnectedness" or "color dimension" do not seem to originate directly from sensory data. I'm wondering where these live. In your model everything seems connected to sensors.

As always feel free to ignore the questions. I know I'm being difficult, but you seem to be a thinker, perhaps you enjoy a little food for thought.

Louis Savain said...

Peter, these are great thoughts. I'll respond with my own thoughts sometime later this week, after the festivities.

Louis Savain said...

Hi Peter, I am rather busy with mundane but important stuff. Your questions are important. I don't have all the answers but when I get some free time, I will write a full article to address some of them.

Peter ( said...

No worries Louis. Life goes on, that is good. I'll keep an eye on here, take care.