A gradient-based machine learning optimization algorithm that computes individual adaptive learning rates for each parameter, combining the advantages of Adagrad and RMSprop.