1: \begin{abstract}

2:   This paper presents an asynchronous incremental aggregated gradient algorithm

3:   and its implementation in a parameter server framework for solving regularized

4:   optimization problems. The algorithm can handle both general convex (possibly

5:   non-smooth) regularizers and general convex constraints. When the empirical

6:   data loss is strongly convex, we establish linear convergence rate, give

7:   explicit expressions for step-size choices that guarantee convergence to the

8:   optimum, and bound the associated convergence factors. The expressions have an

9:   explicit dependence on the degree of asynchrony and recover classical results

10:   under synchronous operation. Simulations and implementations on commercial

11:   compute clouds validate our findings.

12: \end{abstract}

13: