A method used to facilitate training of deep learning neural network machine learning models by optimizing and finding the best network parameters (including bias and weights) to minimize the model’s cost function and thereby get to the desired results more quickly.