Templated graphical models (TGMs) encode model structure using rules that capture recurring relationships between multiple random variables. While the rules in TGMs are interpretable, it is not clear how they can be used to generate explanations for the individual predictions of the model. Further, learning these rules from data comes with high computational costs: it typically requires an expensive combinatorial search over the space of rules and repeated optimization over rule weights. In this work, we propose a new structure learning algorithm, Explainable Structured Model Search (ESMS), that learns a templated graphical model and an explanation framework for its predictions. ESMS uses a novel search procedure to efficiently search the space of models and discover models that trade-off predictive accuracy and explainability. We introduce the notion of relational stability and prove that our proposed explanation framework is stable. Further, our proposed piecewise pseudolikelihood (PPLL) objective does not require re-optimizing the rule weights across models during each iteration of the search. In our empirical evaluation on three realworld datasets, we show that our proposed approach not only discovers models that are explainable, but also significantly outperforms existing state-out-the-art structure learning approaches.

Abstract Statistical relational learning (SRL) and graph neural networks (GNNs) are two powerful approaches for learning and inference over graphs. Typically, they are evaluated in terms of simple metrics such as accuracy over individual node labels. Complexaggregate graph queries(AGQ) involving multiple nodes, edges, and labels are common in the graph mining community and are used to estimate important network properties such as social cohesion and influence. While graph mining algorithms support AGQs, they typically do not take into account uncertainty, or when they do, make simplifying assumptions and do not build full probabilistic models. In this paper, we examine the performance of SRL and GNNs on AGQs over graphs with partially observed node labels. We show that, not surprisingly, inferring the unobserved node labels as a first step and then evaluating the queries on the fully observed graph can lead to sub-optimal estimates, and that a better approach is to compute these queries as an expectation under the joint distribution. We propose a sampling framework to tractably compute the expected values of AGQs. Motivated by the analysis of subgroup cohesion in social networks, we propose a suite of AGQs that estimate the community structure in graphs. In our empirical evaluation, we show that by estimating these queries as an expectation, SRL-based approaches yield up to a 50-fold reduction in average error when compared to existing GNN-based approaches.

Various industries are turning to social media to identify key influencers on topics of interest. Following this trend, the All England Lawn Tennis and Croquet Club (AELTC) is keen to analyze the `social pulse' around the famous Wimbledon Championships. IBM developed and deployed social influence analysis capability for AELTC during the 2014 edition of the Championship. The design and implementation of influence analysis technology in the real world involves several challenges. In this paper, we define various functional and usability criteria that social influence scores should satisfy, and propose a multi-dimensional definition of influence that satisfies these criteria. We highlight the need to identify both all-time influencers and recent influencers, and track user influences over multiple time-scales for this purpose. We also stress the importance of aspect-specific influence analysis, and investigate an approach that uses an aspect hierarchy that annotates tweets with topics or aspects before analyzing them for influence. We also describe interesting insights discovered by our tool and the lessons that we learnt from this engagement.

The analysis of network connections, diffusion processes and cascades requires evaluating properties of the diffusion network. Properties of interest often involve variables that are not explicitly observed in real world diffusions. Connection strengths in the network and diffusion paths of infections over the network are examples of such hidden variables. These hidden variables therefore need to be estimated for these properties to be evaluated. In this paper, we propose and study this novel problem in a Bayesian framework by capturing the posterior distribution of these hidden variables given the observed cascades, and computing the expectation of these properties under this posterior distribution. We identify and characterize interesting network diffusion properties whose expectations can be computed exactly and efficiently, either wholly or in part. For properties that are not `nice' in this sense, we propose a Gibbs Sampling framework for Monte Carlo integration. In detailed experiments using various network diffusion properties over multiple synthetic and real datasets, we demonstrate that the proposed approach is significantly more accurate than a frequentist plug-in baseline. We also propose a map-reduce implementation of our framework and demonstrate that this can analyze cascades with millions of infections in minutes.