All terms
Foundations
Activation
The intermediate value a neuron or layer computes as data passes through a network.
Definition
An activation is the intermediate value a neural network computes at a unit or layer as input flows forward. Each neuron takes a weighted sum of its inputs, adds a bias, and applies an activation function to produce this number, which then feeds the next layer. Activations are the running representations a model builds up, and inspecting them is central to interpretability.