ABSTRACT
Hebbian associative learning is a common form of neuronal adaptation in the brain. In this paper [1] we show that a Hebbian synapse is an ideal neuronal substrate for adaptive control and other supervised learning tasks. In particular, a homosynaptic Hebbian synapse that exhibits pairing-specific and activity dependent long-term potentiation (LTP) may constitute an adaptive element in a high-gain adaptive control (HGAC) system. Similarly, an associative Hebbian synapse that is conditioned by a heterosynaptic pathway is capable of implementing the delta adaptation rule that is widely used in model-reference adaptive control (MRAC) and error-backpropagation learning schemes.
