1: \begin{abstract}

2: A stochastic incremental subgradient algorithm 

3: for the minimization of a sum of convex functions is

4: introduced. The method sequentially uses partial subgradient

5: information and the sequence of partial subgradients is

6: determined by a general Markov chain. This makes it suitable

7: to be used in networks where the path of information flow is

8: stochastically selected. We prove convergence of the

9: algorithm to a weighted objective function where the weights

10: are given by the \rafaelc{Cesàro} limiting probability distribution 

11: of the Markov chain. Unlike previous works in the literature, the Cesàro

12: limiting distribution is general (not necessarily uniform),

13: allowing for general weighted objective functions and

14: flexibility in the method.

15: \linebreak

16: \textbf{keywords:} Optimization algorithms, incremental subgradient method, stochastic optimization, Markov chains.

17: \end{abstract}

18: