Weights are the parameters of a model that neurons use to decide whether to activate (fire). Their optimal values are learned through backpropagation during the training process. Once learned, these weights can be serialized and deployed for making inferences.
