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.

Several kernel based methods for multi-task learning have been proposed, which leverage relations among tasks as regularization to enhance the overall learning accuracies. These methods assume that the tasks share the same kernel, which could limit their applications because in practice different tasks may need different kernels. The main challenge of introducing multiple kernels into multiple tasks is that models from different Reproducing Kernel Hilbert Spaces (RKHSs) are not comparable, making it difficult to exploit relations among tasks. This paper addresses the challenge by formalizing the problem in the Square Integrable Space (SIS). Specially, it proposes a kernel based method which makes use of a regularization term defined in the SIS to represent task relations. We prove a new representer theorem for the proposed approach in SIS. We further derive a practical method for solving the learning problem and conduct consistency analysis of the method. We discuss the relations between our method and an existing method. We also give an SVM based implementation of our method for multi-label classification. Experiments on two real-world data sets show that the proposed method performs better than the existing method.

We propose a fast batch learning method for linear-chain Conditional Random Fields (CRFs) based on Newton-CG methods. Newton-CG methods are a variant of Newton method for high-dimensional problems. They only require the Hessian-vector products instead of the full Hessian matrices. To speed up Newton-CG methods for the CRF learning, we derive a novel dynamic programming procedure for the Hessian-vector products of the CRF objective function. The proposed procedure can reuse the byproducts of the time-consuming gradient computation for the Hessian-vector products to drastically reduce the total computation time of the Newton-CG methods. In experiments with tasks in natural language processing, the proposed method outperforms a conventional quasi-Newton method. Remarkably, the proposed method is competitive with online learning algorithms that are fast but unstable.

The Approximate Linear Programming (ALP) approach to value function approximation for MDPs is a parametric value function approximation method, in that it represents the value function as a linear combination of features which are chosen a priori. Choosing these features can be a difficult challenge in itself. One recent effort, Regularized Approximate Linear Programming (RALP), uses L1 regularization to address this issue by combining a large initial set of features with a regularization penalty that favors a smooth value function with few non-zero weights. Rather than using smoothness as a backhanded way of addressing the feature selection problem, this paper starts with smoothness and develops a non-parametric approach to ALP that is consistent with the smoothness assumption. We show that this new approach has some favorable practical and analytical properties in comparison to (R)ALP.

Applications that adapt to a particular end user often make inaccurate predictions during the early stages when training data is limited. Although an end user can improve the learning algorithm by labeling more training data, this process is time consuming and too ad hoc to target a particular area of inaccuracy. To solve this problem, we propose a new learning algorithm based on Locally Weighted Logistic Regression for feature labeling by end users, enabling them to point out which features are important for a class, rather than provide new training instances. In our user study, the first allowing ordinary end users to freely choose features to label directly from text documents, our algorithm was more effective than others at leveraging end users’ feature labels to improve the learning algorithm. Our results strongly suggest that allowing users to freely choose features to label is a promising method for allowing end users to improve learning algorithms effectively.

We describe the Fourier basis, a linear value function approximation scheme based on the Fourier series. We empirically demonstrate that it performs well compared to radial basis functions and the polynomial basis, the two most popular fixed bases for linear value function approximation, and is competitive with learned proto-value functions.

Multi-task learning has received increasing attention in the past decade. Many supervised multi-task learning methods have been proposed, while unsupervised multi-task learning is still a rarely studied problem. In this paper, we propose to learn a kernel for multi-task clustering. Our goal is to learn a Reproducing Kernel Hilbert Space, in which the geometric structure of the data in each task is preserved, while the data distributions of any two tasks are as close as possible. This is formulated as a unified kernel learning framework, under which we study two types of kernel learning: nonparametric kernel learning and spectral kernel design. Both types of kernel learning can be solved by linear programming. Experiments on several cross-domain text data sets demonstrate that kernel k-means on the learned kernel can achieve better clustering results than traditional single-task clustering methods. It also outperforms the newly proposed multi-task clustering method.

The problem of influence maximization, i.e., mining top-k influential nodes from a social network such that the spread of influence in the network is maximized, is NP-hard. Most of the existing algorithms for the prob- lem are based on greedy algorithm. Although greedy algorithm can achieve a good approximation, it is computational expensive. In this paper, we propose a totally different approach based on Simulated Annealing(SA) for the influence maximization problem. This is the first SA based algorithm for the problem. Additionally, we propose two heuristic methods to accelerate the con- vergence process of SA, and a new method of comput- ing influence to speed up the proposed algorithm. Experimental results on four real networks show that the proposed algorithms run faster than the state-of-the-art greedy algorithm by 2-3 orders of magnitude while being able to improve the accuracy of greedy algorithm.

In many real world applications, the number of examples to learn from is plentiful, but we can only obtain limited information on each individual example. We study the possibilities of efficient, provably correct, large-scale learning in such settings. The main theme we would like to establish is that large amounts of examples can compensate for the lack of full information on each individual example. The type of partial information we consider can be due to inherent noise or from constraints on the type of interaction with the data source. In particular, we describe and analyze algorithms for budgeted learning, in which the learner can only view a few attributes of each training example, and algorithms for learning kernel-based predictors, when individual examples are corrupted by random noise.

With smart-phones becoming increasingly commonplace, there has been a subsequent surge in applications that continuously track the location of users. However, serious privacy concerns arise as people start to widely adopt these applications. Users will need to maintain policies to determine under which circumstances to share their location. Specifying these policies however, is a cumbersome task, suggesting that machine learning might be helpful. In this paper, we present a user-controllable method for learning location sharing policies. We use a classifier based on multivariate Gaussian mixtures that is suitably modified so as to restrict the evolution of the underlying policy to favor incremental and therefore human-understandable changes as new data arrives. We evaluate the model on real location-sharing policies collected from a live location-sharing social network, and we show that our method can learn policies in a user-controllable setting that are just as accurate as policies that do not evolve incrementally. Additionally, we highlight the strength of the generative modeling approach we take, by showing how our model easily extends to the semi-supervised setting.