0

When a neural network is created for a certain memory, what makes that neural network relevant to the memory? In other words, what property of the neural network stimulates the retrieval of the memory?

  • 1
    Does this answer your question? – David Cian Dec 13 '21 at 23:42
  • "An attractor network is a graph like ours, which invariably tends toward an attractor. A simple model of memory that is an attractor network is the Hopfield network. Once retrieval has been initiated, the population of neurons will eventually settle on a single memory."

    No, this doesn't answer my question. I am looking for "What is the connection between a memory and a neural pathway", not "How does the memory settle in a neural pathway". Thanks anyways!

     –  Dec 14 '21 at 00:15
  • 1
    Can you maybe elaborate a bit on what you mean by "what property of the neural network stimulates the retrieval of memory"? Memory retrieval is initiated externally by setting the initial state of neurons in a neuron assembly to something slightly similar to the actual memory, and then the assembly converges to the fully accurate memory: such a system is called autoassociative memory. – David Cian Dec 14 '21 at 01:26
  • I think I didn't phrase it clearly. Let's say we have a neural network that stores the memory of the taste of orange juice. If I damage one neuron in that pathway, and then replace it with a new one, will this pathway still store the taste of orange juice? If not, what caused it to change? –  Dec 14 '21 at 02:23
  • You can think of it like a lock and key but sloppier. In this many-factor approximation, the more of the pertaining input neurons activated (and less of the wrong neurons), the more likely and more certain the memory is to activate. Sub-threshold would put the memory or mental object into perceptive set while crossing the threshold would bring the object into conscious awareness. – Michael Dec 14 '21 at 03:19

0 Answers0