An overview of machine learning techniques in constraint solving

Abstract

Constraint solving is applied in different application contexts. Examples thereof are the configuration of complex products and services, the determination of production schedules, and the determination of recommendations in online sales scenarios. Constraint solvers apply, for example, search heuristics to assure adequate runtime performance and prediction quality. Several approaches have already been developed showing that machine learning (ML) can be used to optimize search processes in constraint solving. In this article, we provide an overview of the state of the art in applying ML approaches to constraint solving problems including constraint satisfaction, SAT solving, answer set programming (ASP) and applications thereof such as configuration, constraint-based recommendation, and model-based diagnosis. We compare and discuss the advantages and disadvantages of these approaches and point out relevant directions for future work.

Knowledge-based recommender systems: overview and research directions

“…In the example shown in Table 12 , variable value orderings for the not yet instantiated variables ABtesting and license would be ABtesting [1,0] and license [100,0], indicating that the solver should first try to instantiate these variables with 1 (100) before trying other instantiations. For an overview on the integration of machine learning and constraint solving, we refer to Popescu et al ( 2022 ).…”

“…Optimizing search efficiency is often based on the application of machine learning. A detailed overview of integration scenarios for constraint solving and machine learning is given in the study by Popescu et al ( 2022 ). Specifically in the context of constraint-based recommendation, multi-criteria optimization becomes an issue since (1) constraint-based recommenders are typically applied in interactive scenarios with the need of efficient solution search and (2) at the same time, solutions must be personalized, i.e., take into account the preferences of a user.…”

“…A major disadvantage is “setup” efforts that are needed to predefine the recommendation knowledge in terms of product properties, product catalogs, similarity metrics, and constraints. There are many new developments in knowledge-based (recommender) systems focusing on a “deep” integration with machine learning , thus helping to unify the two worlds (Popescu et al, 2022 ). However, there are still a couple of research directions that are of interest—see the following discussion.…”

“…Specifically in constraint-based recommender systems, a major line of research focuses on the integration of concepts and techniques from machine learning into constraint-based reasoning, for example, by learning variable and variable value ordering heuristics (Popescu et al, 2022 ; Uta et al, 2022 ). This way, recommendation functionalities can be directly integrated with constraint solving.…”

INFORMEDQX: Informed Conflict Detection for Over-Constrained Problems

“…Especially in interactive settings, there is often a need to identify preferred conflicts (Junker 2004;O'Sullivan et al 2007;Walsh 2007;Rossi, Venable, and Walsh 2011), i.e., conflicts whose resolution could be regarded as acceptable for a user. For example, users of a car configurator with strong preferences regarding an upper price limit are more inclined (in the case that a configurator cannot find a solu-tion) to accept some relaxations of technical car features before accepting to further extend the pre-defined price limit.…”