Non-Asymptotic Analysis of Stochastic Approximation Algorithms for Machine Learning

We consider the minimization of a convex objective function defined on a Hilbert space, which is only available through unbiased estimates of its gradients. This problem includes standard machine lear