[Previous] [Next]

Storing Patterns in a Hopfield Network

• The Hopfield network is designed to store a number of patterns so that they can be retrieved from noisy or partial cues.
• It does this by creating an energy surface which has attractors representing each of the patterns.
• The noisy and partial cues are states of the system which are close to the attractors. As a Hopfield network cycles it slides from the noisy pattern down the energy surface into the closest attractor - representing the closest stored pattern.
• One of the more visually appealing demonstrations involves storing a set of images. The network can then be presented with either a portion of one of the images (partial cue) or an image degraded with noise (noisy cue) and through multiple iterations it will attempt to reconstruct one of the stored images.