Translations:Gradient Descent/1/en

gradient descent is a first-order iterative optimisation algorithm for finding a local minimum of a differentiable function. It is the foundation of nearly all modern machine-learning training procedures, from simple linear regression to billion-parameter deep neural networks.