Catálogo de publicaciones - libros

Many networks are important because they are substrates for dynamical systems, and their pattern of functional connectivity can itself be dynamic — they can functionally reorganize, even if their underlying anatomical structure remains fixed. However, the recent rapid progress in discovering the community structure of networks has overwhelmingly focused on that constant anatomical connectivity. In this paper, we lay out the problem of discovering functional communities , and describe an approach to doing so. This method combines recent work on measuring information sharing across stochastic networks with an existing and successful community-discovery algorithm for weighted networks. We illustrate it with an application to a large biophysical model of the transition from beta to gamma rhythms in the hippocampus.

Social networks are the focus of a large body of research. A number of popular email encryption tools make use of online directories to store public key information. These can be used to build a social network of people connected by email relationships. Since these directories contain creation and expiration time-stamps, the corresponding network can be built and analysed dynamically. At any given point, a snapshot of the current state of the model can be observed and traditional metrics evaluated and compared with the state of the model at other times. We show that, with this described data set, simple traditional predictive measures do vary with time. Moreover, singular events pertinent to the participants in the social network (such as conferences) can be correlated with or implied by significant changes in these measures. This provides evidence that the dynamic behaviour of social networks should not be ignored, either when analysing a real model or when attempting to generate a synthetic model.

This paper surveys work from the field of machine learning on the problem of within-network learning and inference. To give motivation and context to the rest of the survey, we start by presenting some (published) applications of within-network inference. After a brief formulation of this problem and a discussion of probabilistic inference in arbitrary networks, we survey machine learning work applied to networked data, along with some important predecessors—mostly from the statistics and pattern recognition literature. We then describe an application of within-network inference in the domain of suspicion scoring in social networks. We close the paper with pointers to toolkits and benchmark data sets used in machine learning research on classification in network data. We hope that such a survey will be a useful resource to workshop participants, and perhaps will be complemented by others.

Online social networks are often a by-product of blogging and other online media sites on the Internet. Services such as LiveJournal allow their users to specify who their “friends” are, and thus a social network is formed. Some users choose not to disclose personal information which their friends list. This paper will explore the relationship between users with the intent of being able to make a prediction of a users age and country of residence based on the information given by their friends on this social network.

There is increasing interest in the storage, retrieval, and analysis of email communications. One active area of research focuses on the inference of properties of the underlying social network giving rise to the email communications[1,2]. Email communication between individuals implies some type of relationship, whether it is formal, such as a manager-employee relationship, or informal, such as friendship relationships. Understanding the nature of these observed relationships can be problematic given there is a shared context among the individuals that isn’t necessarily communicated. This provides a challenge for analysts that wish to explore and understand email archives for legal or historical research.

In this abstract we address the problem of learning approximate Markov Random Fields (MRF) from large transactional data. Examples of such data include market basket data, co-authorship networked data, etc. Such data can be represented by a binary data matrix, with an entry ( i, j ) takes a value of one (zero) if the item j is (not) in the basket i . “Large” means that there can be many rows or columns in the data matrix. To model such data effectively in order to answer queries about the data efficiently, we consider the use of probabilistic models. In this abstract, we consider employing frequent itemsets to learn approximate global MRFs on large transactional data. We conduct an empirical study on real datasets to show the efficiency and effectiveness of our model on solving the query selectivity estimation problem, that is to approximately compute the marginal probability of sets of items (see [1] for the experimental results). Translated into the social network domain, this is the problem of computing the likelihood of seeing a particular combination of grocery items in the market basket domain, or the probability of a group of professors coauthoring a paper in a co-authorship network, etc. This marginal probability computation is also useful for anomalous link detection [2] in social network analysis. A link in a social network corresponds to a pair of items. The links whose associated marginal probabilities are significantly low can be thought of as anomalous.

In this volume, we have seen several compelling reasons for the statistical analysis of network data. 1 Find statistical regularities in an observed set of relationships between objects. For example, what kinds of patterns are there in the friendships between co-workers? 1 Understand and make predictions about the specific behavior of certain actors in a domain. For example, who is Jane likely to be friends with given the friendships we know about? 1 Make predictions about a new actor, having observed other actors and their relationships. For example, when someone new moves to town, what can we predict about his or her relationships to others? 1 Use network data to make predictions about an actor-specific variable. For example, can we predict the functions of a set of proteins given only the protein-protein interaction data? All of the analysis techniques proposed here are model-based: one defines an underlying joint probability distribution on graphs and considers the observed relationship data under that distribution. Loosely—and this will be a point of discussion among the panelists—the models can be divided into those that are “descriptive” or “discriminative” and those that are “generative.”