Mathematics Colloquia and Seminars

Return to Colloquia & Seminar listing

BBPH - Using Progressive Hedging inside Branch and Bound nodes to Solve Multi-Stage Stochastic Mixed Integer Programs

Student-Run Applied & Math Seminar

Speaker: Jason Barnett, UC Davis
Location: 2112 MSB
Start time: Wed, Jan 20 2016, 12:10PM

Progressive hedging (PH), though an effective heuristic for stochastic MIPs, is not guaranteed convergence in the integer case. Here, we propose BBPH, a branch and bound algorithm that uses PH within each node that, given enough time, will always converge to the optimal solution. Interesting avenues of research emerge answering the questions of obtaining bounds in a reasonable amount of time, exploiting parallelism, and branching decisions.