In this paper, we study the problem of learning ametric and propose a loss function based metriclearning framework, in which the metric is estimatedby minimizing an empirical risk over a trainingset. With mild conditions on the instance distributionand the used loss function, we prove that theempirical risk converges to its expected counterpartat rate O(1/\sqrt{n}), wherein n is the cardinality of the training set. In addition, with the assumption thatthe best metric that minimizes the expected risk isbounded, we prove that the learned metric is consistent. Two example algorithms are presented by usingthe proposed loss function based metric learningframework, each of which uses a log loss functionand a smoothed hinge loss function, respectively. Experimental results on data sets from the UCI machine learning repository suggest the effectivenessof the proposed algorithms.

Topic models have a wide range of applications, including modeling of text documents, images, user preferences, product rankings, and many others. However, learning optimal models may be difficult, especially for large problems. The reason is that inference techniques such as Gibbs sampling often converge to suboptimal models due to the abundance of local minima in large datasets. In this paper, we propose a general method of improving the performance of topic models. The method, called 'grouping transform', works by introducing auxiliary variables which represent assignments of the original model tokens to groups. Using these auxiliary variables, it becomes possible to resample an entire group of tokens at a time. This allows the sampler to make larger state space moves. As a result, better models are learned and performance is improved. The proposed ideas are illustrated on several topic models and several text and image datasets. We show that the grouping transform significantly improves performance over standard models.

Heuristics used for solving hard real-time search problems have regions with depressions. Such regions are bounded areas of the search space in which the heuristic function is exceedingly low compared to the actual cost to reach a solution. Real-time search algorithms easily become trapped in those regions since the heuristic values of states in them may need to be updated multiple times, which results in costly solutions. State-of-the-art real-time search algorithms like LSS-LRTA*, LRTA*(k), etc., improve LRTA*'s mechanism to update the heuristic, resulting in improved performance. Those algorithms, however, do not guide search towards avoiding or escaping depressed regions. This paper presents depression avoidance, a simple real-time search principle to guide search towards avoiding states that have been marked as part of a heuristic depression. We apply the principle to LSS-LRTA* producing aLSS-LRTA*, a new real-time search algorithm whose search is guided towards exiting regions with heuristic depressions. We show our algorithm outperforms LSS-LRTA* in standard real-time benchmarks. In addition we prove aLSS-LRTA* has most of the good theoretical properties of LSS-LRTA*.

This problem involves imprecise probability judgements, widening the scope of application areas. In fact, there is a In this paper, we study algorithms for probabilistic large number of potential application areas for PSAT, from satisfiability (PSAT), an NPcomplete problem, machine learning to the modelling of biological processes, and their empiric complexity distribution. We define from hardware and software verification to economics and a PSAT normal form, based on which we propose econometrics. However, there are very few, if any, practical two logic-based algorithms: a reduction of algorithms available, used in limited applications.

One approach for artificially intelligent agents wishing to maximise some performance metric in a given domain is to learn from a collection of training data that consists of actions or decisions made by some expert, in an attempt to imitate that expert's style. We refer to this type of agent as an expert imitator. In this paper we investigate whether performance can be improved by combining decisions from multiple expert imitators. In particular, we investigate two existing approaches for combining decisions. The first approach combines decisions by employing ensemble voting between multiple expert imitators. The second approach dynamically selects the best imitator to use at runtime given the performance of the imitators in the current environment. We investigate these approaches in the domain of computer poker. In particular, we create expert imitators for limit and no limit Texas Hold'em and determine whether their performance can be improved by combining their decisions using the two approaches listed above.

How do we scale information extraction to the massive size and unprecedented heterogeneity of the Web corpus? Beginning in 2003, our KnowItAll project has sought to extract high-quality knowledge from the Web. In 2007, we introduced the Open Information Extraction (Open IE) paradigm which eschews handlabeled training examples, and avoids domain-specific verbs and nouns, to develop unlexicalized, domain-independent extractors that scale to the Web corpus. Open IE systems have extracted billions of assertions as the basis for both common-sense knowledge and novel question-answering systems. This paper describes the second generation of Open IE systems, which rely on a novel model of how relations and their arguments are expressed in English sentences to double precision/recall compared with previous systems such as TEXTRUNNER and WOE.

Purely data driven approaches for machine learning present difficulties when data is scarce relative to the complexity of the model or when the model is forced to extrapolate. On the other hand, purely mechanistic approaches need to identify and specify all the interactions in the problem at hand (which may not be feasible) and still leave the issue of how to parameterize the system. In this paper, we present a hybrid approach using Gaussian processes and differential equations to combine data driven modelling with a physical model of the system. We show how different, physically-inspired, kernel functions can be developed through sensible, simple, mechanistic assumptions about the underlying system. The versatility of our approach is illustrated with three case studies from motion capture, computational biology and geostatistics.

Twitter and other microblogs have rapidly become a significant means by which people communicate with the world and each other in near realtime. There has been a large number of studies surrounding these social media, focusing on areas such as information spread, various centrality measures, topic detection and more. However, one area which has not received much attention is trying to better understand what information is being spread and why it is being spread. This work looks to get a better understanding of what makes people spread information in tweets or microblogs through the use of retweeting. Several retweet behavior models are presented and evaluated on a Twitter data set consisting of over 768,000 tweets gathered from monitoring over 30,000 users for a period of one month. We evaluate the proposed models against each user and show how people use different retweet behavior models. For example, we find that although users in the majority of cases do not retweet information on topics that they themselves Tweet about as or from people who are "like them" (hence anti-homophily), we do find that models which do take homophily, or similarity, into account fits the observed retweet behaviors much better than other more general models which do not take this into account. We further find that, not surprisingly, people's retweeting behavior is better explained through multiple different models rather than one model.

As online social networks expand their role beyond maintaining existing relationships, they may look to more faceted ratings to support the formation of new connections between their users. Our study focuses on one community employing faceted ratings, CouchSurfing.org, and combines data analysis of ratings, a large-scale survey, and in-depth interviews. In order to understand the ratings, we revisit the notions of friendship and trust and uncover an asymmetry: close friendship includes trust, but high levels of trust can be achieved without close friendship. To users, providing faceted ratings presents challenges, including differentiating and quantifying inherently subjective feelings such as friendship and trust, concern over a friend's reaction to a rating, and knowledge of how ratings can affect others' reputations. One consequence of these issues is the near absence of negative feedback, even though a small portion of actual experiences and privately held ratings are negative. We show how users take this into account when formulating and interpreting ratings, and discuss designs that could encourage more balanced feedback.

To experiment properly, scientists from many researchareas need large sets of real world data. Information re-trieval scientists for example often need to evaluate theiralgorithms on a dataset or a gold standard. The availabil-ity of these datasets often is insufficient and authors withthe same goal do not evaluate their approaches on thesame data. To make research results more transparentand comparable, we introduce Areca, an online portalfor sharing datasets and/or the results that were reachedwith the author’s algorithms on these datasets. Havingsuch an online comparison makes it easier to grasp thestate-of-the-art on certain tasks and drive research toimprove the results.