A Hopfield network is a totally connected neural network often used as an autoencoder. Like Boltzmann machine, the Hopfield network is related to spin glass models in physics. Hopfield networks differ from Boltzmann machines principally in the update model for node weights, notably, while Hopfield networks adopt a continuous learning mechanism with smooth thresholds, similar to that in backpropagation; Boltzmann machine use stochastic firing that effectively creates a Markov process, and are more similar to spiking neural networks.
Used in glossary entries: autoencoder, backpropagation, Boltzmann machine, neural network, spiking neural network, spin glass models
Links:
neuronaldynamics.epfl.ch: Neuronal Dynamics: Hopfield Model- Wikipedia: Hopfield network
- arXiv: Hopfield Networks is All You Need