ConceptLearning

// Core-mechanism: ConceptSynapseModel#Hebbian (The adaption of weights of synapses).

// The adjustment of synaptic weights is a constant cleanup process: useful connections are strengthened, while irrelevant or incorrect connections are weakened or removed entirely. This saves energy and increases processing speed.

// One could create a model in which synapses are only strengthened but never weakened. Disadvantage: The system would quickly become saturated with overly strong connections, would be unable to “unlearn” anything, and would no longer be able to recognize subtle differences.

// One might assume that learning is controlled by an external “teacher” (as in many AI models, e.g., backpropagation). Disadvantage: This is less biologically plausible than the local Hebbian rule, in which each synapse “learns” for itself.