Yuri Malitsky

Algorithm selection and scheduling

Algorithm portfolios aim to increase the robustness of our ability to solve problems efficiently. While recently proposed algorithm selection methods come ever closer to identifying the most appropriate solver given an input instance, they are bound to make wrong and, at times, costly decisions. Solver scheduling has been proposed to boost the performance of algorithm selection. Scheduling tries to allocate time slots to the given solvers in a portfolio so as to maximize, say, the number of solved instances within a given time limit. We show how to solve the corresponding optimization problem at a low computational cost using column generation, resulting in fast and high quality solutions. We integrate this approach with a recently introduced algorithm selector, which we also extend using other techniques. We propose various static as well as dynamic scheduling strategies, and demonstrate that in comparison to pure algorithm selection, our novel combination of scheduling and solver selection can significantly boost performance.

Instance-Specific Algorithm Configuration

This book presents a modular and expandable technique in the rapidly emerging research area of automatic configuration and selection of the best algorithm for the instance at hand. The author presents the basic model behind ISAC and then details a number of modifications and practical applications. In particular, he addresses automated feature generation, offline algorithm configuration for portfolio generation, algorithm selection, adaptive solvers, online tuning, and parallelization. The authors related thesis was honorably mentioned (runner-up) for the ACP Dissertation Award in 2014, and this book includes some expanded sections and notes on recent developments. Additionally, the techniques described in this book have beensuccessfullyapplied to a number of solvers competing in the SATand MaxSAT International Competitions, winning a total of 18 gold medals between 2011 and 2014. The book will be of interest to researchers and practitioners in artificial intelligence, in particular in the area of machine learning and constraint programming.

Proteus: A Hierarchical Portfolio of Solvers and Transformations

In recent years, portfolio approaches to solving SAT problems and CSPs have become increasingly common. There are also a number of different encodings for representing CSPs as SAT instances. In this paper, we leverage advances in both SAT and CSP solving to present a novel hierarchical portfolio-based approach to CSP solving, which we call Proteus, that does not rely purely on CSP solvers. Instead, it may decide that it is best to encode a CSP problem instance into SAT, selecting an appropriate encoding and a corresponding SAT solver. Our experimental evaluation used an instance of Proteus that involved four CSP solvers, three SAT encodings, and six SAT solvers, evaluated on the most challenging problem instances from the CSP solver competitions, involving global and intensional constraints. We show that significant performance improvements can be achieved by Proteus obtained by exploiting alternative view-points and solvers for combinatorial problem-solving.

Instance-Specific algorithm configuration as a method for non-model-based portfolio generation

Instance-specific algorithm configuration generalizes both instance-oblivious algorithm tuning as well as algorithm portfolio generation. ISAC is a recently proposed non-model-based approach for tuning solver parameters dependent on the specific instance that needs to be solved. While ISAC has been compared with instance-oblivious algorithm tuning systems before, to date a comparison with portfolio generators and other instance-specific algorithm configurators is crucially missing. In this paper, among others, we provide a comparison with SATzilla, as well as three other algorithm configurators: Hydra, DCM and ArgoSmart. Our experimental comparison shows that non-model-based ISAC significantly outperforms prior state-of-the-art algorithm selectors and configurators. The following study was the foundation for the best sequential portfolio at the 2011 SAT Competition.

The 2014 International Planning Competition: Progress and Trends

We review the 2014 International Planning Competition (IPC-2014), the eighth in a series of competitions starting in 1998. IPC-2014 was held in three separate parts to assess state-of-the-art in three prominent areas of planning research: the deterministic (classical) part (IPCD), the learning part (IPCL), and the probabilistic part (IPPC). Each part evaluated planning systems in ways that pushed the edge of existing planner performance by introducing new challenges, novel tasks, or both. The competition surpassed again the number of competitors than its predecessor, highlighting the competition’s central role in shaping the landscape of ongoing developments in evaluating planning systems.

Tuning Parameters of Large Neighborhood Search for the Machine Reassignment Problem

Data centers are a critical and ubiquitous resource for providing infrastructure for banking, Internet and electronic commerce. One way of managing data centers efficiently is to minimize a cost function that takes into account the load of the machines, the balance among a set of available resources of the machines, and the costs of moving processes while respecting a set of constraints. This problem is called the machine reassignment problem. An instance of this online problem can have several tens of thousands of processes. Therefore, the challenge is to solve a very large sized instance in a very limited time. In this paper, we describe a constraint programming-based Large Neighborhood Search (LNS) approach for solving this problem. The values of the parameters of the LNS can have a significant impact on the performance of LNS when solving an instance. We, therefore, employ the Instance Specific Algorithm Configuration (ISAC) methodology, where a clustering of the instances is maintained in an offline phase and the parameters of the LNS are automatically tuned for each cluster. When a new instance arrives, the values of the parameters of the closest cluster are used for solving the instance in the online phase. Results confirm that our CP-based LNS approach, with high quality parameter settings, finds good quality solutions for very large sized instances in very limited time. Our results also significantly outperform the hand-tuned settings of the parameters selected by a human expert which were used in the runner-up entry in the 2012 EURO/ROADEF Challenge.

Parallel SAT solver selection and scheduling

Combining differing solution approaches by means of solver portfolios has proven as a highly effective technique for boosting solver performance. We consider the problem of generating parallel SAT solver portfolios. Our approach is based on a recently introduced sequential SAT solver portfolio that excelled at the last SAT competition. We show how the approach can be generalized for the parallel case, and how obstacles like parallel SAT solvers and symmetries induced by identical processors can be overcome. We compare different ways of computing parallel solver portfolios with the best performing parallel SAT approaches to date. Extensive experimental results show that the developed methodology very significantly improves our current parallel SAT solving capabilities.

Non-model-based algorithm portfolios for SAT

When tackling a computationally challenging combinatorial problem, one often observes that some solution approaches work well on some instances, while other approaches work better on other instances. This observation has given rise to the idea of building algorithm portfolios [5]. Leyton-Brown et al. [1], for instance, proposed to select one of the algorithms in the portfolio based on some features of the instance to be solved. This approach has been blessed with tremendous success in the past. Especially in SAT, the SATzilla portfolios [7] have performed extremely well in past SAT Competitions [6].

Feature Filtering for Instance-Specific Algorithm Configuration

Instance-Specific Algorithm Configuration (ISAC) is a novel general technique for automatically generating and tuning algorithm portfolios. The approach has been very successful in practice, but up to now it has been committed to using all the features it was provided. However, traditional feature filtering techniques are not applicable, requiring multiple computationally expensive tuning steps during the evaluation stage. To this end, we show three new evaluation functions that use precomputed runtimes of a collection of untuned solvers to quickly evaluate subsets of features. One of our proposed functions even shows how to generate such an effective collection of solvers when only one highly parameterized solver is available. Using these new functions, we show that the number of features used by ISAC can be reduced to less than a quarter of the original number while often providing significant performance gains. We present numerical results on both SAT and CP domains.

Backdoors to Combinatorial Optimization: Feasibility and Optimality

There has been considerable interest in the identification of structural properties of combinatorial problems that lead, directly or indirectly, to the development of efficient algorithms for solving them. One such concept is that of a backdoor set--a set of variables such that once they are instantiated, the remaining problem simplifies to a tractable form. While backdoor sets were originally defined to capture structure in decision problems with discrete variables, here we introduce a notion of backdoors that captures structure in optimization problems, which often have both discrete and continuous variables. We show that finding a feasible solution and proving optimality are characterized by backdoors of different kinds and size. Surprisingly, in certain mixed integer programming problems, proving optimality involves a smaller backdoor set than finding the optimal solution. We also show extensive results on the number of backdoors of various sizes in optimization problems. Overall, this work demonstrates that backdoors, appropriately generalized, are also effective in capturing problem structure in optimization problems.