Rate coding in knowledge valuation

From supermemo.guru
Jump to navigation Jump to search

This article by Dr Piotr Wozniak is part of SuperMemo Guru series on memory, learning, creativity, and problem solving.

Rate coding of value

In a concept network, for a neuron or a concept map to impart knowledge valuation, it either needs to fire concurrently with value-encoding neurons, or encode value in the output signal. Firing frequency (or firing rate) seems like an attractive candidate for expressing value associated with a concept (represented by a concept neuron or a concept map).

Evidence

Rate coding may have a role in expressing value in a concept network. Here are some clues that come from various areas of neurophysiology:

  • muscles contract in proportion to the frequency of firing (the fact, discovered in 1926, gave birth to the concept of rate coding)
  • greater activation of the neuron increases firing
  • firing frequency may need to reach a critical value for target neurons to get activated
  • place cells show increased frequency of firing in a set location
  • frequency of the signal corresponds with the degree of pain
  • sound pitch is rate coded
  • object color is rate coded
  • rate codes are used to express the position of joints
  • gaze fixation is rate coded and may express the value of the observed object (source)
  • opioids may increase firing of dopamine cells via the presynaptic inhibition of GABA release (source)
  • rate codes are used by CA1 cells to represent time delays (source)

Value decision tree

In the knowledge valuation network, the wave of spreading activation would be modulated by the rate coding of the participating concepts. This would be analogous to expected value estimates in a decision tree. In addition, activations might be subject to a value threshold, which would make the valuation process highly selective and computationally cheap. This would be analogous to cutting off the branches of the decision tree (branch and bound).

Example

Consider encountering a book about bird taxonomy by an enthusiast of ornithology. Once the brain could decode the title of the book, the bird concept might fire up and propagate value in the network. However, a true firestorm might depend on the accessibility of the book. The same book shown on a TV screen will have an entirely different effect as compared to the book unwrapped as a Christmas gift. The book on TV may provide a mild anticipation-based "wanting" reward. The Christmas gift will quickly lead to a high level "liking" reward. In the knowledge valuation network, the value difference may be expressed by a single concept associated with "accessibility".

In a similar fashion, food anticipation reward may instantly be amplified, if the brain succeeds in computing the route of access that signals "obtainability".

In terms of rate coding, the output of the knowledge valuation network would determine the firing frequency of the decision concept. The firing would be enhanced by "obtainability". Alternatively, the "book about birds" concept might increase its firing rate by feedback from the "obtainability" concept. A mental experiment seems to indicate that the latter option might be pretty plausible. When the obtainability is withdrawn, the appetite for the book will clearly not be reduced to the baseline (e.g. as determined by modest firing in response to a book on TV).



For more texts on memory, learning, sleep, creativity, and problem solving, see Super Memory Guru