1: \begin{abstract}

2: This paper considers an inexact primal-dual algorithm for semi-infinite

3: programming (SIP) for which it provides general error bounds. To implement

4: the dual variable update, we create a new prox function for nonnegative

5: measures which turns out to be a generalization of the Kullback-Leibler

6: divergence for probability distributions. We show that under suitable

7: conditions on the error, this algorithm achieves an $\mathcal{O}(1/\sqrt{K})$

8: rate of convergence in terms of the optimality gap and constraint

9: violation. We then use our general error bounds to analyze the convergence

10: and sample complexity of a specific primal-dual SIP algorithm based

11: on Monte Carlo integration. Finally, we provide numerical experiments

12: to demonstrate the performance of our algorithm. 

13: \end{abstract}