This note considers how to modify BnB-ADOPT, a well-known algorithm for optimally solving distributed constraint optimization problems, with a double aim: (i) to avoid sending most of the redundant messages and (ii) to handle cost functions of any arity. Some of the messages exchanged by BnB-ADOPT turned out to be redundant. Removing most of the redundant messages increases substantially communication efficiency: the number of exchanged messages is -in most casesat least three times fewer (keeping the other measures almost unchanged), and termination and optimality are maintained. On the other hand, handling n-ary cost functions was addressed in the original work, but the presence of thresholds makes their practical usage more complex. Both issues -removing most of the redundant messages and efficiently handling n-ary cost functions-can be combined, producing the new version BnB-ADOPT + . Experimentally, we show the benefits of this version over the original one.

No abstract

Graphical model processing is a central problem in artificial intelligence. The optimization of the combined cost of a network of local cost functions federates a variety of famous problems including CSP, SAT and Max-SAT but also optimization in stochastic variants such as Markov Random Fields and Bayesian networks. Exact solving methods for these problems typically include branch and bound and local inference-based bounds. In this paper we are interested in understanding when and how dynamic programming based optimization can be used to efficiently enforce soft local consistencies on Global Cost Functions, defined as parameterized families of cost functions of unbounded arity. Enforcing local consistencies in cost function networks is performed by applying so-called Equivalence Preserving Transformations (EPTs) to the cost functions. These EPTs may transform global cost functions and make them intractable to optimize. We identify as tractable projection-safe those global cost functions whose optimization is and remains tractable after applying the EPTs used for enforcing arc consistency. We also provide new classes of cost functions that are tractable projection-safe thanks to dynamic programming. We show that dynamic programming can either be directly used inside filtering algorithms, defining polynomially DAG-filterable cost functions, or emulated by arc consistency filtering on a Berge-acyclic network of bounded-arity cost functions, defining Berge-acyclic network-decomposable cost functions. We give examples of such cost functions and we provide a systematic way to define decompositions from existing decomposable global constraints. These two approaches to enforcing consistency in global cost functions are then embedded in a solver for extensive experiments that confirm the feasibility and efficiency of our proposal.✩ This paper is an extended version of [3],

Global constraints are an essential component in the efficiency of centralized constraint programming. We propose to include global constraints in distributed constraint satisfaction and optimization problems (DisCSPs and DCOPs). We detail how this inclusion can be done, considering different representations for global constraints (direct, nested, binary). We explore the relation of global constraints with local consistency (both in the hard and soft cases), in particular for generalized arc consistency (GAC). We provide experimental evidence of the benefits of global constraints on several benchmarks, both for distributed constraint satisfaction and for distributed constraint optimization.

Abstract. In the centralized context, global constraints have been essential for the advancement of constraint reasoning. In this paper we propose to include soft global constraints in distributed constraint optimization problems (DCOPs). Looking for efficiency, we study possible decompositions of global constraints, including the use of extra variables. We extend the distributed search algorithm BnB-ADOPT + to support these representations of global constraints. In addition, we explore the relation of global constraints with soft local consistency in DCOPs, in particular for the generalized soft arc consistency (GAC) level. We include specific propagators for some well-known soft global constraints. Finally, we provide empirical results on several benchmarks.