1: \begin{abstract}

2:   This paper considers a class of real-time decision making problems

3:   to minimize the expected value of a function that depends on a

4:   random variable $\xi$ under an unknown distribution \prob. In this

5:   process, samples of $\xi$ are collected sequentially in real time,

6:   and the decisions are made, using the real-time data, to guarantee

7:   out-of-sample performance. We approach this problem in a

8:   distributionally robust optimization framework and propose a novel

9:   \ODAA for this purpose. This algorithm guarantees the out-of-sample

10:   performance in high probability, and gradually improves the quality

11:   of the data-driven decisions by incorporating the streaming data. We

12:   show that the \ODAA guarantees convergence under the streaming data,

13:   and a criteria for termination of the algorithm after certain number

14:   of data has been collected.

15: \end{abstract}

16: