Pattern-Matching and Invariance

In neuroscience, the process I call pattern-matching is sometimes referred to as an invariant operation.

Yohan John in Quora raises the question How does your brain discover association in the midst of so much disassociation? and suggests the form of the answer. That’s a question that has nagged at me for thirty plus years. His answer is that the brain somehow creates invariant representations of objects and patterns. I can put some further structure on that answer.

What is invariance?

In math and physics, an invariant is a property of a system which remains unchanged under some transformation. Here are three short ways the invariant concept is used in neuroscience:

  • Sensory invariance. We can identify an object irrespective of its physical position in relation to us. A table seen from five feet straight on and from twenty feet away at an angle is still recognized as the same table.
  • Partial information. A quick glimpse of the side of face across a crowded dance floor can be enough to identify a friend.
  • Contradictory information. Your friend meets you, but surprise, he has no hair. Nonetheless, you recognize your friend, despite never having seen that look before.

Two Neural Properties

A narrow, but vital thread to the invariant question arises from two properties of the neuron—the neural threshold and the All-or-None Principle.

James Kalat combines both in this statement from his text, Biological Psychology (p 242).

Any stimulation beyond the threshold, regardless of how far beyond, produces the same response.

Together the properties imply that similar, but different neural inputs are treated as if they are identical. If they exceed the neural threshold, they transmit the same signal downstream. This is an inherent abstraction of information by neural processing.

This is clearly related to partial information being recognized as matching more complete information. It also bears directly on the handling of contradictory data. If the difference is slight compared to the totality of features seen, it will be overlooked. That is, they will be treated the same because their neural inputs both surmount the threshold.

Almost Gate

For this reason, I call this neural feature, the Almost Gate. If two inputs are almost the same, downstream they are the same. The two inputs can even be  from separate sources, one originating in current sensory information and the other pulled from memory.

Figure 3.13 All waves higher than the seawall wet the surfaces beyond it.

Figure 3.13 All waves higher than the seawall wet the surfaces beyond it.

If the neuron’s electrical potential exceeds the threshold, it doesn’t matter how much it exceeds it, it goes over the Almost Gate and is treated as identical downstream.

A Pro and a Con Feature

  • The Almost Gate allows (guarantees) that we occasionally make mistakes like jumping to a conclusion which is false. That is we treat two things as identical even though they are not. Two things that are different in a small but significant way are treated as the same. This happens at high levels like conversations, as well as low levels like mistaking a cat’s meow for a child’s cry.
  • Yet, when a person has a relatively low Almost Gate, more patterns cross the threshold. As they appear as equal in later thoughts, this allows for novel connections, some of which are productive. That is a crucial source of creativity.

Seawall photo: www.davis.abwe.org/image/detail/155516.jpg

  3 comments for “Pattern-Matching and Invariance

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload the CAPTCHA.

This site uses Akismet to reduce spam. Learn how your comment data is processed.