1: \begin{abstract}

2: Supported by the recent contributions in multiple branches, the first-order splitting algorithms became central for structured nonsmooth optimization. In the large-scale or noisy contexts, when only stochastic information on the objective function is available, the extension of proximal gradient schemes to stochastic oracles is heavily based on the tractability of the proximal operator corresponding to nonsmooth component, which has been deeply analyzed in the literature. However, there remained some questions about the difficulty of the composite models where the nonsmooth term is not proximally tractable anymore. Therefore, in this paper we tackle composite optimization problems, where the access only to stochastic information on both smooth and nonsmooth components is assumed, using a stochastic proximal first-order scheme with stochastic proximal updates. We provide sublinear $\mathcal{O}\left( \frac{1}{k} \right)$ convergence rates (in expectation of squared distance to the optimal set) under the strong convexity assumption on the objective function. Also, linear convergence is achieved for convex feasibility problems. The empirical behavior is illustrated by numerical tests on parametric sparse representation models.

3: \end{abstract}

4: