We study the application of min-max propagation, a variation of belief propagation, for approximate min-max inference in factor graphs. We show that for "any" high-order function that can be minimized in O(ω), the min-max message update can be obtained using an efficient O(K(ω + log(K)) procedure, where K is the number of variables. We demonstrate how this generic procedure, in combination with efficient updates for a family of high-order constraints, enables the application of min-max propagation to efficiently approximate the NP-hard problem of makespan minimization, which seeks to distribute a set of tasks on machines, such that the worst case load is minimized.

We show that for "any" high-order function that can be minimized in

POST-REBUTTAL UPDATE I have read and considered the authors' response and stand by my initial decision. I would heavily stress that the revised paper more clearly state prior work on message passing specific to the min-max semiring, including Aji and McEliece and Vinyals et al. SUMMARY: The authors propose a dynamic programming algorithm for solving min-max problems with objectives that decompose as a set of lower dimensional factors. For cases where factors have high arity the authors further show an efficient updating scheme in cases where factors can be efficiently minimized with some variables clamped. Further speedups are shown for cases where the domain is constrained to a feasible subset. PROS: This reviewer found the paper to be well-rounded with a clear presentation of algorithmic components and with experiments on an interesting load balancing application.

We study the application of min-max propagation, a variation of belief propagation, for approximate min-max inference in factor graphs. We show that for "any" highorder function that can be minimized in O(ω), the min-max message update can be obtained using an efficient O(K(ω + log(K)) procedure, where K is the number of variables. We demonstrate how this generic procedure, in combination with efficient updates for a family of high-order constraints, enables the application of min-max propagation to efficiently approximate the NP-hard problem of makespan minimization, which seeks to distribute a set of tasks on machines, such that the worst case load is minimized.

We study the application of min-max propagation, a variation of belief propagation, for approximate min-max inference in factor graphs. We show that for "any" high-order function that can be minimized in O(ω), the min-max message update can be obtained using an efficient O(K(ω log(K)) procedure, where K is the number of variables. We demonstrate how this generic procedure, in combination with efficient updates for a family of high-order constraints, enables the application of min-max propagation to efficiently approximate the NP-hard problem of makespan minimization, which seeks to distribute a set of tasks on machines, such that the worst case load is minimized. Papers published at the Neural Information Processing Systems Conference.

We study the application of min-max propagation, a variation of belief propagation, for approximate min-max inference in factor graphs. We show that for “any” high-order function that can be minimized in O(ω), the min-max message update can be obtained using an efficient O(K(ω + log(K)) procedure, where K is the number of variables. We demonstrate how this generic procedure, in combination with efficient updates for a family of high-order constraints, enables the application of min-max propagation to efficiently approximate the NP-hard problem of makespan minimization, which seeks to distribute a set of tasks on machines, such that the worst case load is minimized.