Understanding Hebbian Synapses: Pattern recognition in spiking neurons

Hebbian Synapses are a lot more interesting and complex than just “Neurons which fire together, wire together.” This video shows how synapse weights need to be distributed in order to recognize input patterns with some amount of error tolerance. In this case, the system is recognizing various color values, but the same process could be applied to any sensory input.

There are limitations and trade-offs in synapse-based pattern-recognition. Primarily, the more error the system needs to tolerate, the slower it will run. On the other hand, since additional neurons are needed to recognize additional different patterns, there is no speed penalty for more data—a network which recognizes 5 patterns from 15 inputs (as in the example) will run at the same speed 1,000 patterns from 10,000 inputs.