Perturbed Linear Programming with Application to Markov Decision Processes
(R Burachik, V Gaitsgory and J Filar with K Avrachenkov)
This project involves collaboration between two research institutions, which are UniSA and INRIA, France. The participants are R Burachik (UniSA), V. Gaitsgory (UniSA), J. Filar (UniSA) and K. Avrachenkov (INRIA-France).
We study the limiting properties of a family of LP-problems parameterized by e>0. We look for the limiting optimal values and limiting solution set when e goes to zero, when the problem does not verify Slater conditions.