Real-time heuristic search algorithms that precompute search space-specific databases have demonstrated exceptional performance in video-game pathfinding. We discuss the first steps towards extending these algorithms to other search spaces that also benefit from the real-time property. We present our initial progress in characterizing the performance of current algorithms based on the features of a search space, and discuss future directions of this research.

Dots-And-Boxes is a well-known and widely-played combinatorial game. While the rules of play are very simple, the state space for even small games is extremely large, and finding the outcome under optimal play is correspondingly hard. In this paper we introduce a Dots-And-Boxes solver which is significantly faster than the current state-of-the-art: over an order-of-magnitude faster on several large problems. We describe our approach, which uses Alpha-Beta search and applies a number of techniques—both problem-specific and general—to reduce the number of duplicate states explored and reduce the search space to a manageable size. Using these techniques, we have determined for the first time that Dots- And-Boxes on a board of 4x5 boxes is a tie given optimal play. This is the largest game solved to date.

A major difficulty in a search-based problem-solving process is the task of searching the potentially huge search space resulting from the exponential growth of states. State explosion rapidly occupies memory and increases computation time. Although various heuristic search algorithms have been developed to solve problems in a reasonable time, there is no efficient method to construct heuristic functions. In this work, we propose a method by which a neural network can be iteratively trained to form an efficient heuristic function. An adaptive heuristic search procedure is involved in the training iterations. This procedure reduces the evaluation values of the states that are involved in the currently known best solution paths. By doing so, the promising states are continuously moved forward. The adapted heuristic values are fed back to neural networks; thus, a well-trained network function can find the near-best solutions quickly. To demonstrate this method, we solved the fifteen-puzzle problem. Experimental results showed that the solutions obtained by our method were very close to the shortest path, and both the number of explored nodes and the search time were significantly reduced.

Local search algorithms applied to optimization problems often suffer from getting trapped in a local optimum. The common solution for this deficiency is to restart the algorithm when no progress is observed. Alternatively, one can start multiple instances of a local search algorithm, and allocate computational resources (in particular, processing time) to the instances depending on their behavior. Hence, a multi-start strategy has to decide (dynamically) when to allocate additional resources to a particular instance and when to start new instances. In this paper we propose multi-start strategies motivated by works on multi-armed bandit problems and Lipschitz optimization with an unknown constant. The strategies continuously estimate the potential performance of each algorithm instance by supposing a convergence rate of the local search algorithm up to an unknown constant, and in every phase allocate resources to those instances that could converge to the optimum for a particular range of the constant. Asymptotic bounds are given on the performance of the strategies. In particular, we prove that at most a quadratic increase in the number of times the target function is evaluated is needed to achieve the performance of a local search algorithm started from the attraction region of the optimum. Experiments are provided using SPSA (Simultaneous Perturbation Stochastic Approximation) and k-means as local search algorithms, and the results indicate that the proposed strategies work well in practice, and, in all cases studied, need only logarithmically more evaluations of the target function as opposed to the theoretically suggested quadratic increase.

Automating the design of heuristic search methods is an active research field within computer science, artificial intelligence and operational research. In order to make these methods more generally applicable, it is important to eliminate or reduce the role of the human expert in the process of designing an effective methodology to solve a given computational search problem. Researchers developing such methodologies are often constrained on the number of problem domains on which to test their adaptive, self-configuring algorithms; which can be explained by the inherent difficulty of implementing their corresponding domain specific software components. This paper presents HyFlex, a software framework for the development of cross-domain search methodologies. The framework features a common software interface for dealing with different combinatorial optimisation problems, and provides the algorithm components that are problem specific. In this way, the algorithm designer does not require a detailed knowledge the problem domains, and thus can concentrate his/her efforts in designing adaptive general-purpose heuristic search algorithms. Four hard combinatorial problems are fully implemented (maximum satisfiability, one dimensional bin packing, permutation flow shop and personnel scheduling), each containing a varied set of instance data (including real-world industrial applications) and an extensive set of problem specific heuristics and search operators. The framework forms the basis for the first International Cross-domain Heuristic Search Challenge (CHeSC), and it is currently in use by the international research community. In summary, HyFlex represents a valuable new benchmark of heuristic search generality, with which adaptive cross-domain algorithms are being easily developed, and reliably compared.

ADOPT and BnB-ADOPT are two optimal DCOP search algorithms that are similar except for their search strategies: the former uses best-first search and the latter uses depth-first branch-and-bound search. In this paper, we present a new algorithm, called ADOPT( k ), that generalizes them. Its behavior depends on the k parameter. It behaves like ADOPT when k = 1, like BnB-ADOPT when k = ∞ and like a hybrid of ADOPT and BnB-ADOPT when 1 < k < ∞. We prove that ADOPT( k ) is a correct and complete algorithm and experimentally show that ADOPT( k ) outperforms ADOPT and BnB-ADOPT on several benchmarks across several metrics.

We propose a new additive decomposition of probability tables that preserves equivalence of the joint distribution while reducing the size of potentials, without extra variables. We formulate the Most Probable Explanation (MPE) problem in belief networks as a Weighted Constraint Satisfaction Problem (WCSP). Our pairwise decomposition allows to replace a cost function with smaller-arity functions. The resulting pairwise decomposed WCSP is then easier to solve using state-of-the-art WCSP techniques. Although testing pairwise decomposition is equivalent to testing pairwise independence in the original belief network, we show how to efficiently test and enforce it, even in the presence of hard constraints. Furthermore, we infer additional information from the resulting nonbinary cost functions by projecting and subtracting them on binary functions. We observed huge improvements by preprocessing with pairwise decomposition and project&subtract compared to the current state-of-the-art solvers on two difficult sets of benchmark.

Planning under uncertainty for multiagent systems can be formalized as a decentralized partially observable Markov decision process. We advance the state of the art for optimal solution of this model, building on the Multiagent A* heuristic search method. A key insight is that we can avoid the full expansion of a search node that generates a number of children that is doubly exponential in the node's depth. Instead, we incrementally expand the children only when a next child might have the highest heuristic value. We target a subsequent bottleneck by introducing a more memory-efficient representation for our heuristic functions. Proof is given that the resulting algorithm is correct and experiments demonstrate a significant speedup over the state of the art, allowing for optimal solutions over longer horizons for many benchmark problems.

Transposition tables are a powerful tool in search domains for avoiding duplicate effort and for guiding node expansions. Traditionally, however, they have only been applicable when the current state is exactly the same as a previously explored state. We consider a generalized transposition table, whereby a similarity metric that exploits local structure is used to compare the current state with a neighbourhood of previously seen states. We illustrate this concept and forward pruning based on function approximation in the domain of Skat, and show that we can achieve speedups of 16+ over standard methods.

We develop a real-time algorithm based on a Monte-Carlo game tree search for solving a quantified constraint satisfaction problem (QCSP), which is a CSP where some variables are universally quantified. A universally quantified variable represents a choice of nature or an adversary. The goal of a QCSP is to make a robust plan against an adversary. However, obtaining a complete plan off-line is intractable when the size of the problem becomes large. Thus, we need to develop a real-time algorithm that sequentially selects a promising value at each deadline. Such a problem has been considered in the field of game tree search. In a standard game tree search algorithm, developing a good static evaluation function is crucial. However, developing a good static evaluation function for a QCSP is very difficult since it must estimate the possibility that a partially assigned QCSP is solvable. Thus, we apply a Monte-Carlo game tree search technique called UCT. However, the simple application of the UCT algorithm does not work since the player and the adversary are asymmetric, i.e., finding a game sequence where the player wins is very rare. We overcome this difficulty by introducing constraint propagation techniques. We experimentally compare the winning probability of our UCT-based algorithm and the state-of-the-art alpha-beta search algorithm. Our results show that our algorithm outperforms the state-of-the-art algorithm in large-scale problems.