Rational Convex Programs, Their Feasibility, and the Arrow-Debreu Nash Bargaining Game (1010.4280v1)

Abstract: Over the last decade, combinatorial algorithms have been obtained for exactly solving several nonlinear convex programs. We first provide a formal context to this activity by introducing the notion of {\em rational convex programs} -- this also enables us to identify a number of questions for further study. So far, such algorithms were obtained for total problems only. Our main contribution is developing the methodology for handling non-total problems, i.e., their associated convex programs may be infeasible for certain settings of the parameters. The specific problem we study pertains to a Nash bargaining game, called ADNB, which is derived from the linear case of the Arrow-Debreu market model. We reduce this game to computing an equilibrium in a new market model called {\em flexible budget market}, and we obtain primal-dual algorithms for determining feasibility, as well as giving a proof of infeasibility and finding an equilibrium. We give an application of our combinatorial algorithm for ADNB to an important "fair" throughput allocation problem on a wireless channel.