In this paper, we present a supervised learning approach to training submodular scoring functions for extractive multi-document summarization. By taking a structured predicition approach, we provide a large-margin method that directly optimizes a convex relaxation of the desired performance measure. The learning method applies to all submodular summarization methods, and we demonstrate its effectiveness for both pairwise as well as coverage-based scoring functions on multiple datasets. Compared to state-of-the-art functions that were tuned manually, our method significantly improves performance and enables high-fidelity models with numbers of parameters well beyond what could reasonbly be tuned by hand.

We study an approach to policy selection for large relational Markov Decision Processes (MDPs). We consider a variant of approximate policy iteration (API) that replaces the usual value-function learning step with a learning step in policy space. This is advantageous in domains where good policies are easier to represent and learn than the corresponding value functions, which is often the case for the relational MDPs we are interested in. In order to apply API to such problems, we introduce a relational policy language and corresponding learner. In addition, we introduce a new bootstrapping routine for goal-based planning domains, based on random walks. Such bootstrapping is necessary for many large relational MDPs, where reward is extremely sparse, as API is ineffective in such domains when initialized with an uninformed policy. Our experiments show that the resulting system is able to find good policies for a number of classical planning domains and their stochastic variants by solving them as extremely large relational MDPs. The experiments also point to some limitations of our approach, suggesting future work.

Monte Carlo (MC) techniques are often used to estimate integrals of a multivariate function using randomly generated samples of the function. In light of the increasing interest in uncertainty quantification and robust design applications in aerospace engineering, the calculation of expected values of such functions (e.g. performance measures) becomes important. However, MC techniques often suffer from high variance and slow convergence as the number of samples increases. In this paper we present Stacked Monte Carlo (StackMC), a new method for post-processing an existing set of MC samples to improve the associated integral estimate. StackMC is based on the supervised learning techniques of fitting functions and cross validation. It should reduce the variance of any type of Monte Carlo integral estimate (simple sampling, importance sampling, quasi-Monte Carlo, MCMC, etc.) without adding bias. We report on an extensive set of experiments confirming that the StackMC estimate of an integral is more accurate than both the associated unprocessed Monte Carlo estimate and an estimate based on a functional fit to the MC samples. These experiments run over a wide variety of integration spaces, numbers of sample points, dimensions, and fitting functions. In particular, we apply StackMC in estimating the expected value of the fuel burn metric of future commercial aircraft and in estimating sonic boom loudness measures. We compare the efficiency of StackMC with that of more standard methods and show that for negligible additional computational cost significant increases in accuracy are gained.

The ability to plan their own motions and to reliably execute them is an important precondition for autonomous robots. In this paper, we consider the problem of planning the motion of a mobile manipulation robot in the presence of deformable objects in the environment. Our approach combines probabilistic roadmap planning with a deformation simulation system. Since the physical deformation simulation is computationally demanding, we use an efficient variant of Gaussian process regression to estimate the deformation cost for individual objects based on training examples. We generate the training data by employing a simulation system in a preprocessing step. Consequently, no simulations are needed during runtime. We implemented and tested our approach on a mobile manipulation robot. Our experiments show that the robot is able to accurately predict and thus consider the deformation cost its manipulator introduces to the environment during motion planning. Simultaneously, the computation time is substantially reduced compared to a system that performs physical simulations online.

Analyzing chat traffic has important applications for both the military and the civilian world. This paper presents a case study of a real-world application of chat analysis in support of team training exercise in the military. It compares the results of an unsupervised learning approach with those of a supervised classification approach. The paper also discusses some of the specific challenges presented by this domain.

With the growth of online information, many people are challenged in finding and reading the information most important for their interests. From 2008-2010 we built an experimental personalized news system where readers can subscribe to organized channels of information that are curated by experts. AI technology was employed to radically reduce the work load of curators and to efficiently present information to readers. The system has gone through three implementation cycles and processed over 16 million news stories from about 12,000 RSS feeds on over 8000 topics organized by 160 curators for over 600 registered readers. This paper describes the approach, engineering and AI technology of the system.

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.

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.

End-user interactive machine learning is a promising tool for enhancing human productivity and capabilities with large unstructured data sets. Recent work has shown that we can create end-user interactive machine learning systems for specific applications. However, we still lack a generalized understanding of how to design effective end-user interaction with interactive machine learning systems. This work presents three explorations in designing for effective end-user interaction with machine learning in CueFlik, a system developed to support Web image search. These explorations demonstrate that interactions designed to balance the needs of end-users and machine learning algorithms can significantly improve the effectiveness of end-user interactive machine learning.

This paper focuses on analyzing and predicting not-answered questions in Community based Question Answering (CQA) services, such as Yahoo! Answers. In CQA services, users express their information needs by submitting natural language questions and await answers from other human users. Comparing to receiving results from web search engines using keyword queries, CQA users are likely to get more specific answers, because human answerers may catch the main point of the question. However, one of the key problems of this pattern is that sometimes no one helps to give answers, while web search engines hardly fail to response. In this paper, we analyze the not-answered questions and give a first try of predicting whether questions will receive answers. More specifically, we first analyze the questions of Yahoo Answers based on the features selected from different perspectives. Then, we formalize the prediction problem as supervised learning – binary classification problem and leverage the proposed features to make predictions. Extensive experiments are made on 76,251 questions collected from Yahoo! Answers. We analyze the specific characteristics of not-answered questions and try to suggest possible reasons for why a question is not likely to be answered. As for prediction, the experimental results show that classification based on the proposed features outperforms the simple word-based approach significantly.