We investigate the learning rate of multiple kernel leaning (MKL) with elastic-net regularization, which consists of an $\ell_1$-regularizer for inducing the sparsity and an $\ell_2$-regularizer for controlling the smoothness. We focus on a sparse setting where the total number of kernels is large but the number of non-zero components of the ground truth is relatively small, and prove that elastic-net MKL achieves the minimax learning rate on the $\ell_2$-mixed-norm ball. Our bound is sharper than the convergence rates ever shown, and has a property that the smoother the truth is, the faster the convergence rate is.

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.

We investigate and evaluate methods for the characterization of social relations from textual communication context, using e-mail as an example. Social relations are intrinsically characterized by the Cartesian product of weights on various axes (we employ valuation and intensity as examples). The prediction of these characteristics is performed by application of unsupervised learning algorithms on meta-data, communication statistics, and the results of deep linguistic analysis of the message body. Classification of sentiment polarity is chosen as the means of linguistic analysis. We find that prediction accuracy can be improved by introducing limited amounts of additional information.

What will social media sites of tomorrow look like? What behaviors will their interfaces enable? A major challenge for designing new sites that allow a broader range of user actions is the difficulty of extrapolating from experience with current sites without first distinguishing correlations from underlying causal mechanisms. The growing availability of data on user activities provides new opportunities to uncover correlations among user activity, contributed content and the structure of links among users. However, such correlations do not necessarily translate into predictive models. Instead, empirically grounded mechanistic models provide a stronger basis for establishing causal mechanisms and discovering the underlying statistical laws governing social behavior. We describe a statistical physics-based framework for modeling and analyzing social media and illustrate its application to the problems of prediction and inference. We hope these examples will inspire the research community to explore these methods to look for empirically valid causal mechanisms for the observed correlations.

More and more users have been taking various actions to diverse resources referred to by URLs such as news, web pages, images, products, movies as a result of the growth of social media. They are annotating, tweeting in Twitter, reblogging in Tumblr, and Liking in Facebook, etc. Analyses about these diverse actions will be useful for aggregating or integrating diverse resources. In this paper, we view users’ actions to resources as expressing their some interests, and by investigating how their interests are expressed in social media, we get suggestions for aggregations. Our results show that a certain kind of action (such as tagging on Delicious) can be used to make predictions on a different kind of action (such as favorite on Twitter). These analyses will be useful for aggregating or integrating diverse contents on multiple sources. In addition to some experimental analyses, we propose a novel method to predict users’ interests in social media, using time-evolving, multinomial relational data. Our experimental results show that the proposed method significantly outperforms standard tensor analysis and an existing state-of-the-art method (LDA) in prediction tasks.

Growing pool of public-generated bits like online social networking data provides possibility to sense social dynamics in the urban space. In this position paper, we use a location-based online social networking data to sense geo-social activity and analyze the underlying social activity distribution of three different cities: London, Paris, and New York. We find a non-linear distribution of social activity, which follows the Power Law decay function. We perform inter-urban analysis based on social activity distribution and clustering. We believe that our study sheds new light on context-aware urban computing and social sensing.

We consider prominence ranking in graphs involving actors, their artifacts and the artifact groups. When multiple actors contributing to an artifact constitutes a social tie, associations between the artifacts can be used to infer prominence among actors. This is because prominent actors will tend to collaborate on prominent artifacts, and prominent artifacts will be associated with other prominent artifacts. Our testbed example is the DBLP co-authorship graph: multiple authors (the actors) collaborate to publish research papers (the artifacts); collaboration is the social tie. Papers have prominence themselves (eg. quality and impact of the work) and the prominence of the venues are tied to the prominence of the papers in them. We use our methods to infer prominence based on the venue-based associations of papers, and compare our rankings with external citation based measures of prominence. We compare with numerous other ranking algorithms, and show that the ranking performance gain from using the venues is statistically significant. What if there are no natural artifact groups like venues? We develop a new algorithm which uses discovered artifact groups. Our approach consists of two steps. First, we find artifact groups by linking artifacts with common contributors. Note that instead of finding communities of actors, we consider communities of artifacts. We then use these grouped artifacts in the prominence ranking algorithm. We consider different methods for obtaining the artifact groups, in particular a very efficient embedding based algorithm for graph clustering and show the effectiveness of our method in improving the ranking of actors. The inferred groups are as good as or better than the natural conference venues for DBLP.

In August 2010 Facebook launched Places, a location-based service that allows users to check into points of interest and share their physical whereabouts with friends. The friends who see these events in their News Feed can then respond to these check-ins by liking or commenting on them. These data consisting of the places people go and how their friends react to them are a rich, novel dataset. In this paper we first analyze this dataset to understand the factors that influence where users check in, including previous check-ins, similarity to other places, where their friends check in, time of day, and demographics. We show how these factors can be used to build a predictive model of where users will check in next. Then we analyze how users respond to their friends’ check-ins and which factors contribute to users liking or commenting on them. We show how this can be used to improve the ranking of check-in stories, ensuring that users see only the most relevant updates from their friends and ensuring that businesses derive maximum value from check-ins at their establishments. Finally, we construct a model to predict friendship based on check-in count and show that cocheck-ins has a statistically significant effect on friendship.

Another important question is the absoluteness of the constraints. Lately several web-based tagging systems such as Technorati, Even if we use this approach to turn tags into constraints, Flickr or Delicious have become very popular. In this a fair amount of them are bound to be inaccurate paper we will exploit the information created by the community (i.e., linking documents which should not be in the same in Delicious: a social bookmarking service where cluster) until a high value of the parameter t, due to the polysemy the users can save the URLs of their favourite webpages of the terms used as tags or to differences in the criteria offering also the possibility of associating tags to them. of the taggers. Consequently, we have used soft positive On the other hand the clustering methods are a very important constraints, meaning that the documents affected by one of data mining tool in order to exploit the knowledge them are likely to be in the same cluster, without forcing the present in data collections. In the last years a new family of clustering algorithm to actually put them so.

Two-mode networks are a natural representation for many kinds of relational data. These networks are bipartite graphs consisting of two distinct sets ("modes") of entities. For example, one can model multiple recipient email data as a two-mode network of (a) individuals and (b) the emails that they send or receive. In this work we present a statistical model for two-mode network data which posits that individuals belong to latent sets and that the members of a particular set tend to co-appear. We show how to infer these latent sets from observed data using a Markov chain Monte Carlo inference algorithm. We apply the model to the Enron email corpus, using it to discover interpretable latent structure as well as evaluating its predictive accuracy on a missing data task. Extensions to the model are discussed that incorporate additional side information such as the email's sender or text content, further improving the accuracy of the model.