The paper I will talk about is titled ‘Robust Optimization and Network Flows’ and is available here.
This paper deals with the general problem of solving an optimization problem when there is uncertainty in the coefficients of the constraints and the objective function. There are canonically two method for dealing with this: one is Stochastic optimization, where the objective function involves the expectation of the quantity to be maximized. The other is robust optimization, where one tries to optimize a function against the worst possible outcome of the uncertainty.
The authors of this paper give a general procedure for transforming any linear optimization problem (possibly integer) into a linear formulation of the robust counterpart. They also show some nice properties about the size and complexity of the resulting problem, and also give some approximability results that hold under certain assumptions.
I will attempt to go through some of the derivation of the robust formulation that is given in the paper, and also some of the other results such as the approximability.