Showing papers by "Barry O'Sullivan published in 2022"

Finding Counterfactual Explanations through Constraint Relaxations

Abstract: Interactive constraint systems often suffer from infeasibil- ity (no solution) due to conflicting user constraints. A com-mon approach to recover infeasibility is to eliminate the con- straints that cause the conflicts in the system. This approach allows the system to provide an explanation as: “if the user is willing to drop out some of their constraints, there exists a so-lution”. However, one can criticise this form of explanation as not being very informative. A counterfactual explanation is a type of explanation that can provide a basis for the user to re- cover feasibility by helping them understand which changes can be applied to their existing constraints rather than remov- ing them. This approach has been extensively studied in the machine learning field, but requires a more thorough investi- gation in the context of constraint satisfaction. We propose an iterative method based on conflict detection and maximal re- laxations in over-constrained constraint satisfaction problems to help compute a counterfactual explanation.

Generation and Prediction of Difficult Model Counting Instances

Abstract: We present a way to create small yet difficult model counting instances. Our generator is highly parameterizable: the number of variables of the instances it produces, as well as their number of clauses and the number of literals in each clause, can all be set to any value. Our instances have been tested on state of the art model counters, against other difficult model counting instances, in the Model Counting Competition. The smallest unsolved instances of the competition, both in terms of number of variables and number of clauses, were ours. We also observe a peak of difficulty when fixing the number of variables and varying the number of clauses, in both random instances and instances built by our generator. Using these results, we predict the parameter values for which the hardest to count instances will occur.

SATfeatPy - A Python-based Feature Extraction System for Satisfiability

Abstract: Feature extraction is a fundamental task in the application of machine learning methods to SAT solving. It is used in algorithm selection and configuration for solver portfolios and satisfiability classification. Many approaches have been proposed to extract meaningful attributes from CNF instances. Most of them lack a working/updated implementation, and the limited descriptions lack clarity affecting the reproducibility. Furthermore, the literature misses a comparison among the features. This paper introduces SATfeatPy , a library that offers feature extraction techniques for SAT problems in the CNF form. This package offers the implementation of all the structural and statistical features from there major papers in the field. The library is provided in an up-to-date, easy-to-use Python package alongside a detailed feature description. We show the high accuracy of SAT/UNSAT and problem category classification, using five sets of features generated using our library from a dataset of 3000 SAT and UNSAT instances, over ten different classes of problems. Finally, we compare the usefulness of the features and importance for predicting a SAT instance’s original structure in an ablation study.

Variable-Relationship Guided LNS for the Car Sequencing Problem

Abstract: Abstract Large Neighbourhood Search (LNS) is a powerful technique that applies the “divide and conquer” principle to boost the performance of solvers on large scale Combinatorial Optimization Problems. In this paper we consider one of the main hindrances to the LNS popularity, namely the requirement of an expert to define a problem specific neighborhood. We present an approach that learns from problem structure and search performance in order to generate neighbourhoods that can match the performance of domain specific heuristics developed by an expert. Furthermore, we present a new objective function for the optimzation version of the Car Sequencing Problem, that better distinguishes solution quality. Empirical results on public instances demonstrate the effectiveness of our approach against both a domain specific heuristic and state-of-the-art generic approaches.

A Large Neighborhood Search Approach for the Data Centre Machine Reassignment Problem

Abstract: Abstract One of the main challenges in data centre operations involves optimally reassigning running processes to servers in a dynamic setting such that operational performance is improved. In 2012, Google proposed the Machine Reassignment Problem in collaboration with the ROADEF/Euro challenge. A number of complex instances were generated for evaluating the submissions. This work focuses on new approaches to solve this problem. In particular, we propose a Large Neighbourhood Search approach with a novel, domain-specific heuristic for neighborhood selection. This heuristic uses the unbalanced resource usage on the machines to select the most promising processes in each iteration. Furthermore, we compare two search strategies to optimise the sub-problems. The first one is based on the concept of Limited Discrepancy Search, albeit tailored to large scale problems; and the second approach involves the standard combination of constraint programming with random restart strategies. An empirical evaluation on the widely studied instances from ROADEF 2012 demonstrates the effectiveness of our approach against the state-of-the-art, with new upper bounds found for three instances.