An Even More Optimal Stochastic Optimization Algorithm: Minibatching and Interpolation Learning

The massive scale of many modern machine learning models and datasets give rise to complex, high-dimensional training objectives that can be very computationally expensive to optimize.