Yoram Singer: Memory-Efficient Adaptive Optimization for Humungous-Scale Learning

Adaptive gradient-based optimizers such as AdaGrad and Adam are among the methods of choice in modern machine learning. These methods maintain second-order statistics of each model parameter, thus doubling the memory footprint of the optimizer. In behemoth-size applications, this memory overhead restricts the size of the model being used as well as the number of examples in a mini-batch. We describe a novel, simple, and flexible adaptive optimization method with sublinear memory cost that retains the benefits of per-parameter adaptivity while allowing for larger models and mini-batches. We give convergence guarantees for our method and demonstrate its effectiveness in training some of the largest deep models used at Google.