Strategy mining is a new area of research about discovering strategies in decision-making. In this paper, we formulate the strategy-mining problem as a clustering problem, called the latent-strategy problem. In a latent-strategy problem, a corpus of data instances is given, each of which is represented by a set of features and a decision label. The inherent dependency of the decision label on the features is governed by a latent strategy. The objective is to find clusters, each of which contains data instances governed by the same strategy. Existing clustering algorithms are inappropriate to cluster dependency because they either assume feature independency (e.g., K-means) or only consider the co-occurrence of features without explicitly modeling the special dependency of the decision label on other features (e.g., Latent Dirichlet Allocation (LDA)). In this paper, we present a baseline unsupervised learning algorithm for dependency clustering. Our model-based clustering algorithm iterates between an assignment step and a minimization step to learn a mixture of decision tree models that represent latent strategies. Similar to the Expectation Maximization algorithm, our algorithm is grounded in the statistical learning theory. Different from other clustering algorithms, our algorithm is irrelevant-feature resistant and its learned clusters (modeled by decision trees) are strongly interpretable and predictive. We systematically evaluate our algorithm using a common law dataset comprised of actual cases. Experimental results show that our algorithm significantly outperforms K-means and LDA on clustering dependency.

Opponent modeling is a critical mechanism in repeated games. It allows a player to adapt its strategy in order to better respond to the presumed preferences of its opponents. We introduce a modeling technique that adaptively balances safety and exploitability. The opponent's strategy is modeled with a set of possible strategies that contains the actual one with high probability. The algorithm is safe as the expected payoff is above the minimax payoff with high probability, and can exploit the opponent's preferences when sufficient observations are obtained. We apply the algorithm to a robot table-tennis setting where the robot player learns to prepare to return a served ball. By modeling the human players, the robot chooses a forehand, backhand or middle preparation pose before they serve. The learned strategies can exploit the opponent's preferences, leading to a higher rate of successful returns.

In this paper, we present the exact online updating formulae for the generalized inverse of centered matrices. The computational cost is O ( mn ) for matrices of size m × n . Experimental results validate the proposed method’s accuracy and efficiency.  

MAPP has been previously shown as a state-of-the-art multi-agent path planning algorithm on criteria including scalability and success ratio (i.e., percentage of solved units) on realistic game maps. MAPP further provides a formal characterization of problems it can solve, and low-polynomial upper bounds on the resources required. However, until now, MAPP's solution quality had not been extensively analyzed. In this work we empirically analyze the quality of MAPP's solutions, using multiple quality criteria such as the total travel distance, the makespan and the sum of actions (including move and wait actions). We also introduce enhancements that improve MAPP's solution quality significantly. For example, the sum of actions is cut to half on average. The improved MAPP is competitive in terms of solution quality with FAR and WHCA*, two successful algorithms from the literature, and maintains its advantages on different performance criteria, such as scalability, success ratio, and ability to tell apriori if it will succeed in the instance at hand. As optimal algorithms have limited scalability, evaluating the quality of the solutions provided by suboptimal algorithms is another important topic. Using lower bounds of optimal values, we show that MAPP's solutions have a reasonable quality. For example, MAPP's total travel distance is on average 19% longer than a lower bound on the optimal value.

Integrating the expressive power of first-order logic with the ability of probabilistic reasoning of Bayesian networks has attracted the interest of many researchers for decades. We present an approach to integration that translates logical knowledge into Bayesian networks and uses Bayesian network composition to build a uniform representation that supports both logical and probabilistic reasoning. In particular, we propose a new way of translation of logical knowledge, relation search. Through the use of the proposed framework, without learning new languages or tools, modelers are allowed to 1) specify special knowledge using the most suitable languages, while reasoning in a uniform engine; 2) make use of pre-existing logical knowledge bases for probabilistic reasoning (to complete the model or minimize potential inconsistencies).

In this work we present an advanced Bayesian formulation to the task of control learning that employs the Relevance Vector Machines (RVM) generative model for value function evaluation. The key aspect of the proposed method is the design of the discount return as a generalized linear model that constitutes a well-known probabilistic approach. This allows to augment the model with advantageous sparse priors provided by the RVM's regression framework. We have also taken into account the significant issue of selecting the proper parameters of the kernel design matrix. Experiments have shown that our method produces improved performance in both simulated and real test environments.

This work proposes new approaches to contingent planning using alternative belief state representations extended from those in conformant planning and a new AND/OR forward search algorithm, called PrAO, for contingent solutions. Each representation was implemented in a new contingent planner. The important role of belief state representation has been confirmed by the fact that our planners all outperform other stateof- the-art planners on most benchmarks and the comparison of their performances varies across all the benchmarks even using the same search algorithm PrAO and same unsophisticated heuristic scheme. The work identifies the properties of each representation method that affect the performance.

Evolutionary Algorithms (EA) are a branch of heuristic population-based optimization tools that is growing in popularity (especially for combinatorial and other problems with poorly understood landscapes). Despite their many uses, there are no proofs that an EA will always converge to the global optimum for any general problem.

We present an algorithm for compressing pattern databases (PDBs) and a method for fast random access of these com-pressed PDBs. We demonstrate the effectiveness of our technique by compressing two 6-tile sliding-tile PDBs by a factor of 12 and a 7-tile sliding-tile PDB by a factor of 24.

Their information needs are long and popular information needs of the masses. Topic term and highly dynamic - nearly everything about this topic specificity, customizability, and automatically pursuing the is in flux. For these users, information search can be made long term unique information needs of individual users are more effective with knowledge about the field and about the not among the strengths of current main stream search engines types of documents being retrieved. Because the resource (Jansen, Spink, and Saracevic 2000) (Teevan, Dumais, management decisions require judgment about the materials and Horvitz 2005). This gap has inspired web personalization collected, the users require confidentiality and must trust the and collaborative information seeking tools such as sources. Google Alerts and has encouraged topic-specific blogs and Matilda is designed to 1) tailor information collection for podcasts.