A computational process where a neural network is provided with example inputs and expected outputs, and its weights are adjusted in an iterative process to maximize accuracy.