neural network weights

Terms from Artificial Intelligence: humans at the heart of algorithms

The glossary is being gradually proof checked, but may have typos and misspellings.

Each connection in a neural network typically has a weight, which is used to amplify or diminish the way activation of one neuron feeds forward through the connection to the next neuron. If the weight is negative then the connection is inhibitory, that is activation of the first neuron reduces the activation of the next one.
Most neural networks are organised in layers with uni-directional links from each lower-level layer to the next higher-level one. However, some types of network, such as Hopfield networks have weighted connections within a single layer.

Used in glossary entries: hopfield networks, neural network