Activation quantization

Activation quantization refers to reducing the numerical precision of the activation values (also known as feature maps) in a neural network. Activations are the outputs of layers within a neural network.