Skip to content

ConceptSynapseWeightAdaption

DavidFreely edited this page Nov 10, 2025 · 4 revisions

Related: ConceptSynapseModel#Hebbian Synapse Model, ConceptSynapseWeight

"While the ratios of weight increases to decreases are dictated by the desire for stable synapse weights, the absolute amounts of change are dictated by the desire for weights to change as quickly as possible to facilitate rapid learning."

"Increasing the number of possible synapse weight values would require smaller weight changes for each spike. This can be done by changing entries in the lookup table but would result in correspondingly slower learning rates."

WeightAdaptionAmount

"A precise formula for the amount of increase or decrease of biological synapse weights is not known and within the Brain Simulator is controlled by a lookup table which is subject to change. The selection of the model simply selects the lookup table to use. As of this writing, there are three weight control tables.

List of ConceptSynapseWeightAdaptions

(Additional models are anticipated)

Binary

simultaneous firing causes the synapse weight to be set to 1. Firing the target without the source sets the weight to 0.

Hebbian1

Weights range from 0 to 1 such that all single synapses are stable (see below).

Hebbian2

Weights range from -1 to 1. Weights are varied so pattern recognition is learned with an arbitrary number of input spikes. The synapse weight is limited by the number of incoming synapses to the target neuron.

Clone this wiki locally