DOCSLIB.ORG

Learning Variable Ordering Heuristics for Solving Constraint Satisfaction Problems Wen Song, Zhiguang Cao, Jie Zhang, and Andrew Lim

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Learning Variable Ordering Heuristics for Solving Constraint Satisfaction Problems Wen Song, Zhiguang Cao, Jie Zhang, and Andrew Lim 1 Learning Variable Ordering Heuristics for Solving Constraint Satisfaction Problems Wen Song, Zhiguang Cao, Jie Zhang, and Andrew Lim Abstract—Backtracking search algorithms are often used to strategy. The decision of which variable to select next is solve the Constraint Satisfaction Problem (CSP). The efficiency referred to as variable ordering. It is well acknowledged that of backtracking search depends greatly on the variable order- the choice of variable ordering has a critical impact on the ing heuristics. Currently, the most commonly used heuristics are hand-crafted based on expert knowledge. In this paper, efficiency of backtracking search algorithms [12]. However, we propose a deep reinforcement learning based approach to finding the optimal orderings, i.e. those minimize search effort automatically discover new variable ordering heuristics that are (in terms of number of search nodes, total solving time, better adapted for a given class of CSP instances. We show that etc.), is at least as hard as solving the CSP [13]. Therefore, directly optimizing the search cost is hard for bootstrapping, current practice mainly relies on hand-crafted variable ordering and propose to optimize the expected cost of reaching a leaf node in the search tree. To capture the complex relations among heuristics obtained from the experience of human experts, such the variables and constraints, we design a representation scheme as MinDom [14], Dom/Ddeg [15], and impact-based heuristic based on Graph Neural Network that can process CSP instances [16]. Though they are easy to use and widely adopted, they do with different sizes and constraint arities. Experimental results on not have any formal guarantees on the optimality. In addition, random CSP instances show that the learned policies outperform they are designed for solving any CSP instance without consid- classical hand-crafted heuristics in terms of minimizing the search tree size, and can effectively generalize to instances that ering the domain-specific features, which can be exploited to are larger than those used in training. achieve much better efficiency. However, incorporating these additional features requires substantial experience and deep Index Terms—Constraint Satisfaction Problem, variable order- ing, deep reinforcement learning, Graph Neural Network domain knowledge, which are hard to obtain in reality [17]. Recently, Deep Neural Networks (DNNs) have been shown to be promising in learning algorithms for solving NP-hard I. INTRODUCTION problems, such as Traveling Salesman Problem (TSP), Propo- OMBINATORIAL problems [1] are widely studied in sitional Satisfiability Problem (SAT), and Capacitated Vehicle C many research areas, and have numerous real-world ap- Routing Problem (CVRP) [18], [19], [20], [21], [22], [23], plications in domains such as planning and scheduling [2], [3], [24], [25], [26], [27]. The effectiveness comes from the fact vehicle routing [4], [5], graph problems [6], [7], computational that given a class of problem instances (e.g. drawn from a biology [8], [9], etc. As one of the most widely studied distribution), DNN can be trained to discover useful patterns combinatorial problems in computer science and artificial that may not be known or hard to be specified by human intelligence, Constraint Satisfaction Problem (CSP) provides experts, through supervised or reinforcement learning (RL). a general framework for modeling and solving combinatorial In this paper, we ask the following question: can we use problems. A CSP instance involves a set of variables and DNN to discover better variable ordering heuristics for a constraints. To solve it, one needs to find a value assignment class of CSP? This is not a trivial task, due to the following for all variables such that all constraints are satisfied, or prove two challenges. Firstly, given the exponential (worst-case) such assignment does not exist. Despite its ubiquitous appli- complexity of CSP, it is not practical to obtain large amount arXiv:1912.10762v2 [cs.AI] 12 Nov 2020 cations, unfortunately, CSP is well known to be NP-complete of labeled training data (e.g. optimal search paths), therefore it in general [10]. To solve CSP efficiently, backtracking search is hard to apply supervised learning methods. Secondly, CSP algorithms are often employed, which are exact algorithms instances have different sizes and features (e.g. number of vari- with the guarantee that a solution will be found if one exists. ables and constraints, domain of each variable, tightness and Though the worst-case complexity is still exponential, with arity of each constraint). It is crucial to design a representation the help of constraint propagation [11], backtracking search scheme that can effectively process any CSP instance. algorithms often perform reasonably well in practice. To address these challenges, we design a reinforcement In general, a backtracking search algorithm performs depth- learning agent in this paper, which tries to make the optimal first traverse of a search tree, and tries to find a solution by variable ordering decisions at each decision point to minimize iteratively selecting a variable and applying certain branching search cost. Note that here our objective is to minimize the search tree size, which is more suitable for evaluating different W. Song is with the Institute of Marine Science and Technology. Email: [email protected]. ordering heuristics [28]. More specifically, variable ordering in Z. Cao and A. Lim are with the Department of Industrial Systems Engi- backtracking search is modeled as a Markov Decision Process neering and Management, National University of Singapore. Email: isecaoz, (MDP), where the optimal policy is to select at each decision [email protected]. J. Zhang is with the School of Computer Science and Engineering, Nanyang point the variable with the minimum expected search cost. Technological University, Singapore. Email: [email protected] The RL agent can optimize its policy for this MDP by learning 2 from its own experiences of solving CSP instances drawn from respectively). A more promising way is to apply machine a distribution, without the need of supervision. However, such learning within the framework of exact algorithms, such that direct formulation is not convenient for bootstrapping, and the feasibility and solution quality can be guaranteed [38]. A learning must be delayed until backtracking from a search typical exact framework is the branch-and-bound algorithm node. To resolve this issue, we consider the search paths for solving Mixed Integer Linear Programs (MILPs). He et originated from a node as separate trajectories, and opt to al. [39] use imitation learning to learn a control policy for minimize the expected cost of reaching a leaf node. We selecting branches in the branch-and-bound process. Khalil et represent the internal states of the search process based on al. [40] achieves similar purpose by solving a learning-to-rank Graph Neural Network (GNN) [29], which can process CSP task to mimic the behaviors of strong branching. Khalil et al. instances of any size and constraint arity, and effectively [41] also develop a machine learning model to decide whether capture the relationship between the variables and constraints. the primal heuristics should be run for a given branch-and- We use Double Deep Q-Network (DDQN) [30] to train the bound node. These methods are based on linear models with GNN based RL agent. Experimental results on random CSP static and dynamic features describing the current branch-and- instances generated by the well-known model RB [31] show bound status. More recently, Gasse et al. [42] use imitation that the RL agent can discover policies that are better than the learning to mimic strong branching, where the underlying traditional hand-crafted variable ordering heuristics, in terms states are represented using GNN. Similarly, a GNN based of minimizing the search tree size. More importantly, the network is designed in [43], which is trained in a supervised learned policy can effectively generalize to larger instances way to predict values of binary variables in MILP. Though that have never been seen during training. sharing similar GNN structure, our work differs from [42], [43] in mainly two aspects. First, our method does not require labels that are costly to obtain but necessary for imitation or II. RELATED WORK supervised learning. Second, as will be shown latter, we only Recently, there has been an increasing attention on using uses 4 simple raw features, while [42], [43] rely on 19 and 22 deep (reinforcement) learning to tackle hard combinatorial complex MILP features, respectively. (optimization or satisfaction) problems. Quite a few works Another exact framework is the backtracking search algo- in this direction focus on solving specific types of problems, rithms for solving satisfaction problems. Balafrej et al. [44] including routing (e.g. TSP and CVRP) [18], [19], [20], [21], use bandit model to learn a policy that can adaptively select [22], [23], graph optimization [24], [25], packing problems the right constraint propagation levels at each node of a CSP [32], [33], and scheduling [34], [35]. Instead of solving search tree. More close to our work, several methods use tradi- specific problems, we focus on CSP which is a general tional machine learning to choose the branching heuristics for representation of combinatorial problems. solving CSP and some special cases. Lagoudakis and Littman In the literature, a number of methods try to tackle sat- [45] use RL to learn the branching rule selection policy for isfaction problems such as CSP and SAT in an end-to- the #DPLL algorithm for solving SAT, which requires finding end fashion, meaning that training DNN to directly output all solutions for a satisfiable instance. However, as will be a solution for a given instance. Xu et al. [36] represent discussed in Section IV, this RL formulation is not directly binary CSP as a matrix and train a Convolutional Neural applicable for learning in our case. Samulowitz and Memisevic Network (CNN) to predict its satisfiability, but cannot give [46] study the heuristic selection task for solving Quantified the solution for satisfiable instances.
Load more

Recommended publications
  • Constraint Satisfaction with Preferences Ph.D
    MASARYK UNIVERSITY BRNO FACULTY OF INFORMATICS } Û¡¢£¤¥¦§¨ª«¬­Æ !"#°±²³´µ·¸¹º»¼½¾¿45<ÝA| Constraint Satisfaction with Preferences Ph.D. Thesis Brno, January 2001 Hana Rudová ii Acknowledgements I would like to thank my supervisor, Ludˇek Matyska, for his continuous support, guidance, and encouragement. I am very grateful for his help and advice, which have allowed me to develop both as a person and as the avid student I am today. I want to thank to my husband and my family for their support, patience, and love during my PhD study and especially during writing of this thesis. This research was supported by the Universities Development Fund of the Czech Repub- lic under contracts # 0748/1998 and # 0407/1999. Declaration I declare that this thesis was composed by myself, and all presented results are my own, unless otherwise stated. Hana Rudová iii iv Contents 1 Introduction 1 1.1 Thesis Outline ..................................... 1 2 Constraint Satisfaction 3 2.1 Constraint Satisfaction Problem ........................... 3 2.2 Optimization Problem ................................ 4 2.3 Solution Methods ................................... 5 2.4 Constraint Programming ............................... 6 2.4.1 Global Constraints .............................. 7 3 Frameworks 11 3.1 Weighted Constraint Satisfaction .......................... 11 3.2 Probabilistic Constraint Satisfaction ........................ 12 3.2.1 Problem Definition .............................. 13 3.2.2 Problems’ Lattice ............................... 13 3.2.3 Solution
    [Show full text]
  • Backtracking Search (Csps) ■Chapter 5 5.3 Is About Local Search Which Is a Very Useful Idea but We Won’T Cover It in Class
    CSC384: Intro to Artificial Intelligence Backtracking Search (CSPs) ■Chapter 5 5.3 is about local search which is a very useful idea but we won’t cover it in class. 1 Hojjat Ghaderi, University of Toronto Constraint Satisfaction Problems ● The search algorithms we discussed so far had no knowledge of the states representation (black box). ■ For each problem we had to design a new state representation (and embed in it the sub-routines we pass to the search algorithms). ● Instead we can have a general state representation that works well for many different problems. ● We can build then specialized search algorithms that operate efficiently on this general state representation. ● We call the class of problems that can be represented with this specialized representation CSPs---Constraint Satisfaction Problems. ● Techniques for solving CSPs find more practical applications in industry than most other areas of AI. 2 Hojjat Ghaderi, University of Toronto Constraint Satisfaction Problems ●The idea: represent states as a vector of feature values. We have ■ k-features (or variables) ■ Each feature takes a value. Domain of possible values for the variables: height = {short, average, tall}, weight = {light, average, heavy}. ●In CSPs, the problem is to search for a set of values for the features (variables) so that the values satisfy some conditions (constraints). ■ i.e., a goal state specified as conditions on the vector of feature values. 3 Hojjat Ghaderi, University of Toronto Constraint Satisfaction Problems ●Sudoku: ■ 81 variables, each representing the value of a cell. ■ Values: a fixed value for those cells that are already filled in, the values {1-9} for those cells that are empty.
    [Show full text]
  • Prolog Lecture 6
    Prolog lecture 6 ● Solving Sudoku puzzles ● Constraint Logic Programming ● Natural Language Processing Playing Sudoku 2 Make the problem easier 3 We can model this problem in Prolog using list permutations Each row must be a permutation of [1,2,3,4] Each column must be a permutation of [1,2,3,4] Each 2x2 box must be a permutation of [1,2,3,4] 4 Represent the board as a list of lists [[A,B,C,D], [E,F,G,H], [I,J,K,L], [M,N,O,P]] 5 The sudoku predicate is built from simultaneous perm constraints sudoku( [[X11,X12,X13,X14],[X21,X22,X23,X24], [X31,X32,X33,X34],[X41,X42,X43,X44]]) :- %rows perm([X11,X12,X13,X14],[1,2,3,4]), perm([X21,X22,X23,X24],[1,2,3,4]), perm([X31,X32,X33,X34],[1,2,3,4]), perm([X41,X42,X43,X44],[1,2,3,4]), %cols perm([X11,X21,X31,X41],[1,2,3,4]), perm([X12,X22,X32,X42],[1,2,3,4]), perm([X13,X23,X33,X43],[1,2,3,4]), perm([X14,X24,X34,X44],[1,2,3,4]), %boxes perm([X11,X12,X21,X22],[1,2,3,4]), perm([X13,X14,X23,X24],[1,2,3,4]), perm([X31,X32,X41,X42],[1,2,3,4]), perm([X33,X34,X43,X44],[1,2,3,4]). 6 Scale up in the obvious way to 3x3 7 Brute-force is impractically slow There are very many valid grids: 6670903752021072936960 ≈ 6.671 × 1021 Our current approach does not encode the interrelationships between the constraints For more information on Sudoku enumeration: http://www.afjarvis.staff.shef.ac.uk/sudoku/ 8 Prolog programs can be viewed as constraint satisfaction problems Prolog is limited to the single equality constraint: – two terms must unify We can generalise this to include other types of constraint Doing so leads
    [Show full text]
  • The Logic of Satisfaction Constraint
    Artificial Intelligence 58 (1992) 3-20 3 Elsevier ARTINT 948 The logic of constraint satisfaction Alan K. Mackworth* Department of Computer Science, University of British Columbia, Vancouver, BC, Canada V6T 1 W5 Abstract Mackworth, A.K., The logic of constraint satisfaction, Artificial Intelligence 58 (1992) 3-20. The constraint satisfaction problem (CSP) formalization has been a productive tool within Artificial Intelligence and related areas. The finite CSP (FCSP) framework is presented here as a restricted logical calculus within a space of logical representation and reasoning systems. FCSP is formulated in a variety of logical settings: theorem proving in first order predicate calculus, propositional theorem proving (and hence SAT), the Prolog and Datalog approaches, constraint network algorithms, a logical interpreter for networks of constraints, the constraint logic programming (CLP) paradigm and propositional model finding (and hence SAT, again). Several standard, and some not-so-standard, logical methods can therefore be used to solve these problems. By doing this we obtain a specification of the semantics of the common approaches. This synthetic treatment also allows algorithms and results from these disparate areas to be imported, and specialized, to FCSP; the special properties of FCSP are exploited to achieve, for example, completeness and to improve efficiency. It also allows export to the related areas. By casting CSP both as a generalization of FCSP and as a specialization of CLP it is observed that some, but not all, FCSP techniques lift to CSP and thereby to CLP. Various new connections are uncovered, in particular between the proof-finding approaches and the alternative model-finding ap- proaches that have arisen in depiction and diagnosis applications.
    [Show full text]
  • Constraint Satisfaction Problems (Backtracking Search)
    CSC384: Introduction to Artificial Intelligence Constraint Satisfaction Problems (Backtracking Search) • Chapter 6 – 6.1: Formalism – 6.2: Constraint Propagation – 6.3: Backtracking Search for CSP – 6.4 is about local search which is a very useful idea but we won’t cover it in class. Torsten Hahmann, CSC384 Introduction to Artificial Intelligence, University of Toronto, Fall 2011 1 Acknowledgements • Much of the material in the lecture slides comes from Fahiem Bacchus, Sheila McIlraith, and Craig Boutilier. • Some slides come from a tutorial by Andrew Moore via Sonya Allin. • Some slides are modified or unmodified slides provided by Russell and Norvig. Torsten Hahmann, CSC384 Introduction to Artificial Intelligence, University of Toronto, Fall 2011 2 Constraint Satisfaction Problems (CSP) • The search algorithms we discussed so far had no knowledge of the states representation (black box). – For each problem we had to design a new state representation (and embed in it the sub-routines we pass to the search algorithms). • Instead we can have a general state representation that works well for many different problems. • We can then build specialized search algorithms that operate efficiently on this general state representation. • We call the class of problems that can be represented with this specialized representation: CSPs – Constraint Satisfaction Problems. Torsten Hahmann, CSC384 Introduction to Artificial Intelligence, University of Toronto, Fall 2011 3 Constraint Satisfaction Problems (CSP) •The idea: represent states as a vector of feature values. – k-features (or variables) – Each feature takes a value. Each variable has a domain of possible values: • height = {short, average, tall}, • weight = {light, average, heavy} •In CSPs, the problem is to search for a set of values for the features (variables) so that the values satisfy some conditions (constraints).
    [Show full text]
  • Constraint Satisfaction
    Constraint Satisfaction Rina Dechter Department of Computer and Information Science University of California, Irvine Irvine, California, USA 92717 dechter@@ics.uci.edu A constraint satisfaction problem csp de ned over a constraint network consists of a nite set of variables, each asso ciated with a domain of values, and a set of constraints. A solution is an assignmentofavalue to each variable from its domain such that all the constraints are satis ed. Typical constraint satisfaction problems are to determine whether a solution exists, to nd one or all solutions and to nd an optimal solution relative to a given cost function. An example of a constraint satisfaction problem is the well known k -colorability. The problem is to color, if p ossible, a given graph with k colors only, such that anytwo adjacent no des have di erent colors. A constraint satisfaction formulation of this problem asso ciates the no des of the graph with variables, the p ossible colors are their domains and the inequality constraints b etween adjacent no des are the constraints of the problem. Each constraint of a csp may b e expressed as a relation, de ned on some subset of variables, denoting their legal combinations of values. As well, constraints 1 can b e describ ed by mathematical expressions or by computable pro cedures. Another known constraint satisfaction problem is SATis ability; the task of nding the truth assignment to prop ositional variables such that a given set of clauses are satis ed. For example, given the two clauses A _ B _ :C ; :A _ D , the assignmentof false to A, true to B , false to C and false to D , is a satisfying truth value assignment.
    [Show full text]
  • Propositional Satisfiability and Constraint Programming
    Propositional Satisfiability and Constraint Programming: A Comparative Survey LUCAS BORDEAUX, YOUSSEF HAMADI and LINTAO ZHANG Microsoft Research Propositional Satisfiability (SAT) and Constraint Programming (CP) have developed as two relatively in- dependent threads of research cross-fertilizing occasionally. These two approaches to problem solving have a lot in common as evidenced by similar ideas underlying the branch and prune algorithms that are most successful at solving both kinds of problems. They also exhibit differences in the way they are used to state and solve problems since SAT’s approach is, in general, a black-box approach, while CP aims at being tunable and programmable. This survey overviews the two areas in a comparative way, emphasizing the similarities and differences between the two and the points where we feel that one technology can benefit from ideas or experience acquired from the other. Categories and Subject Descriptors: F.4.1 [Mathematical Logic and Formal Languages]: Mathematical Logic—Logic and constraint programming; D.3.3 [Programming Languages]: Language Constructs and Features—Constraints; G.2.1 [Discrete Mathematicals]: Combinatorics—Combinatorial algorithms General Terms: Algorithms Additional Key Words and Phrases: Search, constraint satisfaction, SAT ACM Reference Format: Bordeaux, L., Hamadi, Y., and Zhang, L. 2006. Propositional satisfiability and constraint programming: A comparative survey. ACM Comput. Surv. 38, 4, Article 12 (Dec. 2006), 54 pages. DOI = 10.1145/1177352. 1177354 http://doi.acm.org/10.1145/1177352.1177354 1. INTRODUCTION Propositional satisfiability solving (SAT) and constraint programming (CP) are two automated reasoning technologies that have found considerable industrial applications during the last decades. Both approaches provide generic languages that can be used to express complex (typically NP-complete) problems in areas like hardware verification, configuration, or scheduling.
    [Show full text]
  • Techniques for Efficient Constraint Propagation
    Techniques for Efficient Constraint Propagation MIKAEL Z. LAGERKVIST Licentiate Thesis Stockholm, Sweden, 2008 TRITA-ICT/ECS AVH 08:10 ISSN 1653-6363 KTH ISRN KTH/ICT/ECS AVH-08/10–SE SE-100 44 Stockholm ISBN 978-91-7415-154-1 SWEDEN Akademisk avhandling som med tillstånd av Kungl Tekniska högskolan fram- lägges till offentlig granskning för avläggande av teknologie licentiatexamen fredagen den 21 November 2008 klockan 13.15 i N2, Electrum 3, Kista. © Mikael Z. Lagerkvist, 2008 Tryck: Universitetsservice US AB On two occasions I have been asked, ’Pray, Mr. Babbage, if you put into the machine wrong figures, will the right an- swers come out?’ ... I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question. Passages from the Life of a Philosopher Charles Babbage v Abstract This thesis explores three new techniques for increasing the efficiency of constraint propagation: support for incremental propagation, im- proved representation of constraints, and abstractions to simplify prop- agation. Support for incremental propagation is added to a propagator- centered propagation system by adding a new intermediate layer of abstraction, advisors, that capture the essential aspects of a variable- centered system. Advisors are used to give propagators a detailed view of the dynamic changes between propagator runs. Advisors enable the implementation of optimal algorithms for important constraints such as extensional constraints and Boolean linear in-equations, which is not possible in a propagator-centered system lacking advisors. Using Multivalued Decision Diagrams (MDD) as the representa- tion for extensional constraints is shown to be useful for several rea- sons.
    [Show full text]
  • FOUNDATIONS of CONSTRAINT SATISFACTION Edward Tsang
    FOUNDATIONS OF CONSTRAINT SATISFACTION Edward Tsang Department of Computer Science University of Essex Colchester Essex, UK Copyright 1996 by Edward Tsang All rights reserved. No part of this book may be reproduced in any form by photostat, microfilm, or any other means, without written permission from the author. Copyright 1993-95 by Academic Press Limited This book was first published by Academic Press Limited in 1993 in UK: 24-28 Oval Road, London NW1 7DX USA: Sandiego, CA 92101 ISBN 0-12-701610-4 To Lorna Preface Many problems can be formulated as Constraint Satisfaction Problems (CSPs), although researchers who are untrained in this field sometimes fail to recognize them, and consequently, fail to make use of specialized techniques for solving them. In recent years, constraint satisfaction has come to be seen as the core problem in many applications, for example temporal reasoning, resource allocation, schedul- ing. Its role in logic programming has also been recognized. The importance of con- straint satisfaction is reflected by the abundance of publications made at recent conferences such as IJCAI-89, AAAI-90, ECAI-92 and AAAI-92. A special vol- ume of Artificial Intelligence was also dedicated to constraint reasoning in 1992 (Vol 58, Nos 1-3). The scattering and lack of organization of material in the field of constraint satisfac- tion, and the diversity of terminologies being used in different parts of the literature, make this important topic more difficult to study than is necessary. One of the objectives of this book is to consolidate the results of CSP research so far, and to enable newcomers to the field to study this problem more easily.
    [Show full text]
  • Constraint Logic Programming a Short Tutorial
    Constraint Logic Programming a short tutorial Inês Dutra [email protected] ØØÔ»»ÛÛÛº ÙÔºÔØ»Ò× Departamento de Ciência de Computadores Faculdade de Ciências da Universidade do Porto CRACS & INESC-Porto LA LAI-OIL, June 2010 Outline What is CLP? A little bit of history (motivation) Systems, applications and clients Variable domain CLP by example CRACS-INESC-Porto LA & DCC/FCUP Inês Dutra LAI-OIL, June 2010 2 Constraint Logic Programming What is CLP? the use of a rich and powerful language to model optimization problems (not only...) modelling based on variables, domains and constraints CRACS-INESC-Porto LA & DCC/FCUP Inês Dutra LAI-OIL, June 2010 3 CLP Motivation: 1. to offer a declarative way of modelling constraint satisfaction problems (CSP) 2. to solve 2 limitations in Prolog: Each term in Prolog needs to be explicitly evaluated and is not interpreted (evaluated): · ½ is a term that is not evaluated in Prolog. It is only a syntactic representation. a variable can assume one single value. Uniform computation, but not that powerful: depth-first search, “generate-and-test”. 3. to integrate Artificial Intelligence (AI) and Operations Research (OR) CRACS-INESC-Porto LA & DCC/FCUP Inês Dutra LAI-OIL, June 2010 4 CLP CLP can use Artificial Intelligence (AI) techniques to improve the search: propagation, data-driven computation, “forward checking” and “lookahead”. Applications: planning, scheduling, resource allocation, computer graphics, digital circuit design, fault diagnosis etc. Clients: Michelin and Dassault, French railway SNCF, Swissair, SAS and Cathay Pacific, HK International terminals, Eriksson, British Telecom etc. CRACS-INESC-Porto LA & DCC/FCUP Inês Dutra LAI-OIL, June 2010 5 CLP CLP joins 2 research areas: Introduction of richer and powerful data structures to Logic Programming (e.g.: replace unification by efficient manipulation of constraints and domains).
    [Show full text]
  • Constraint Logic Programming for Qualitative and Quantitative Constraint Satisfaction Problems
    ELSEVIER Decision Support Systems 16 (1996) 67-83 Constraint logic programming for qualitative and quantitative constraint satisfaction problems Ho Geun Lee a,., Ronald M. Lee b, Gang Yu c a Information and System Management Department, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong b Erasmus University Research Institute for Decision and Information Systems (EURIDIS), Erasmus University Rotterdam, P.O. Box 1738, 3000 DR Rotterdam, The Netherlands c Department of Management Science and Information Systems, Graduate School of Business, The University of Texas at Austin, Austin, TX 78712-1175, USA Abstract AI and OR approaches have complementary strengths: AI in domain-specific knowledge representation and OR in efficient mathemat:ical computation. Constraint Logic Programming (CLP), which combines these complementary strengths of the AI and OR approach, is introduced as a new tool to formalize a special class of constraint satisfaction problems that include both qualitative and quantitative constraints. The CLP approach is contrasted with the Mixed Integer Programming (MIP) method from a model-theoretic view. Three relative advantages of CLP over MIP are analyzed: (1) representational economies for domain-specific heuristics, (2) partial solutions, and (3) ease of model revision. A case example of constraint satisfaction problems is implemented by MIP and CLP for comparison of the two approaches. The results exhibit those relative advantages of CLP with computational efficiency comparable to MIP. Keywords: Constraint logic programming; Logic modelling; Constraint solving I. Introduction Since most decisions are made under restrictions or constraints, Constraint Satisfaction Problems (CSP) have been widely studied in both OR and AI. In the OR approach, constraints are quantitative, and a special algorithm like simplex method optimizes single or multiple objective functions subject to numeric constraints.
    [Show full text]
  • Usability Improvements of Optaplanner Benchmarker
    Masarykova univerzita Fakulta}w¡¢£¤¥¦§¨ informatiky !"#$%&'()+,-./012345<yA| Usability improvements of OptaPlanner Benchmarker Diploma thesis Bc. Matej Čimbora Brno, January 2015 Declaration Hereby I declare, that this paper is my original authorial work, which I have worked out by my own. All sources, references and literature used or excerpted during elaboration of this work are properly cited and listed in complete reference to the due source. Bc. Matej Čimbora Advisor: doc. RNDr. Tomáš Pitner, Ph.D. ii Acknowledgement I would like to thank doc. RNDr. Tomáš Pitner, Ph.D. and RNDr. Filip Nguyen for valuable advisory and comments. I owe special thanks to Geoffrey de Smet and Lukáš Petrovický for guidance and help during development process. iii Abstract This thesis is dedicated to improvements of OptaPlanner’s benchmark- ing functionality. Selected usability issues are presented, together with proposed solutions, while the main attention is paid on result persis- tence and comparison. In addition, a short introduction into constraint programming, together with an overview of the main features of Opta- Planner engine is included. iv Keywords OptaPlanner, benchmarking, constraint programming, optimization, Java v Contents 1 Introduction ...........................3 1.1 Goals .............................3 1.2 Structure ..........................4 2 Constraint programming ...................5 3 Configuration and basic usage ...............8 3.1 Modeling the problem ...................8 3.2 Solver configuration ..................... 10 3.2.1 Solution evaluation . 12 3.2.2 Termination . 15 3.3 Running the solver ..................... 15 4 Supported algorithms ..................... 17 4.1 Local search algorithms ................... 17 4.1.1 Hill climbing . 17 4.1.2 Simulated annealing . 18 4.1.3 Tabu search . 18 4.1.4 Late acceptance .
    [Show full text]