Explainable AI (XAI) is an important area but remains relatively understudied for clustering. We propose an explainable-by-design clustering approach that not only finds clusters but also exemplars to explain each cluster. The use of exemplars for understanding is supported by the exemplar-based school of concept definition in psychology. We show that finding a small set of exemplars to explain even a single cluster is computationally intractable; hence, the overall problem is challenging. We develop an approximation algorithm that provides provable performance guarantees with respect to clustering quality as well as the number of exemplars used. This basic algorithm explains all the instances in every cluster whilst another approximation algorithm uses a bounded number of exemplars to allow simpler explanations and provably covers a large fraction of all the instances. Experimental results show that our work is useful in domains involving difficult to understand deep embeddings of images and text.

We consider the setting of cascades that result from contagion dynamics on large realistic contact networks. We address the question of whether the structural properties of a (partially) observed cascade can characterize the contagion scenario and identify the interventions that might be in effect. Using epidemic spread as a concrete example, we study how social interventions such as compliance in social distancing, extent (and efficacy) of vaccination, and the transmissibility of disease can be inferred. The techniques developed are more generally applicable to other contagions as well.

There are myriad real-life examples of contagion processes on human social networks, e.g., spread of viruses, information, and social unrest. Also, there are many methods to control or block contagion spread. In this work, we introduce a novel method of blocking contagions that uses nodes from dominating sets (DSs). To our knowledge, this is the first use of DS nodes to block contagions. Finding minimum dominating sets of graphs is an NP-Complete problem, so we generalize a well-known heuristic, enabling us to customize its execution. Our method produces a prioritized list of dominating nodes, which is, in turn, a prioritized list of blocking nodes. Thus, for a given network, we compute this list of blocking nodes and we use it to block contagions for all blocking node budgets, contagion seed sets, and parameter values of the contagion model. We report on computational experiments of the blocking efficacy of our approach using two mined networks. We also demonstrate the effectiveness of our approach by comparing blocking results with those from the high degree heuristic, which is a common standard in blocking studies.

Fear of reprisals such as violence and punishment can inhibit citizens from speaking out, or make them more reluctant to act, in opposition to a repressive regime. Protests are one form of opposition, and their growth has been successfully modeled as an inftuence-based contagion process within a social network (representing a population). In these models, an individual joins a protest if a sufficient number of her neighbors has already joined. This required number of neighbors is often called a “threshold.” In this study, we model a regime's ability to suppress protests by instilling fear in a subset of a population, and this fear is manifested by an increase in a person's threshold. We consider different social networks, numbers of seed nodes, and amounts of fear. Through simulations, we present several results. For example, we demonstrate that, for the objective of reducing the size of a protest, inducing fear can be more advantageous than removing nodes from a network.

Active learning aims to reduce labeling efforts by selectively asking humans to annotate the most important data points from an unlabeled pool and is an example of human-machine interaction. Though active learning has been extensively researched for classification and ranking problems, it is relatively understudied for regression problems. Most existing active learning for regression methods use the regression function learned at each active learning iteration to select the next informative point to query. This introduces several challenges such as handling noisy labels, parameter uncertainty and overcoming initially biased training data. Instead, we propose a feature-focused approach that formulates both sequential and batch-mode active regression as a novel bipartite graph optimization problem. We conduct experiments on both noise-free and noisy settings. Our experimental results on benchmark data sets demonstrate the effectiveness of our proposed approach.

In anagram games, players are provided with letters for forming as many words as possible over a specified time duration. Anagram games have been used in controlled experiments to study problems such as collective identity, effects of goal-setting, internal-external attributions, test anxiety, and others. The majority of work on anagram games involves individual players. Recently, work has expanded to group anagram games where players cooperate by sharing letters. In this work, we analyze experimental data from online social networked experiments of group anagram games. We develop mechanistic and data-driven models of human decision-making to predict detailed game player actions (e.g., what word to form next). With these results, we develop a composite agent-based modeling and simulation platform that incorporates the models from data analysis. We compare model predictions against experimental data, which enables us to provide explanations of human decision-making and behavior. Finally, we provide illustrative case studies using agent-based simulations to demonstrate the efficacy of models to provide insights that are beyond those from experiments alone.

Abduction is an inference approach that uses data and observations to identify plausible (and preferably, best) explanations for phenomena. Applications of abduction (e.g., robotics, genetics, image understanding) have largely been devoid of human behavior. Here, we devise and execute an iterative abductive analysis process that is driven by the social sciences: behaviors and interactions among groups of human subjects. One goal is to understand intra-group cooperation and its effect on fostering collective identity. We build an online game platform; perform and analyze controlled laboratory experiments; form hypotheses; build, exercise, and evaluate network-based agent-based models; and evaluate the hypotheses in multiple abductive iterations, improving our understanding as the process unfolds. While the experimental results are of interest, the paper's thrust is methodological, and indeed establishes the potential of iterative abductive looping for the (computational) social sciences.

The problem of inferring unknown parameters of a networked social system is of considerable practical importance. We consider this problem for the independent cascade model using an active query framework. More specifically, given a network whose edge probabilities are unknown, the goal is to infer the probability value on each edge by querying the system. The optimization objective is to use as few queries as possible in carrying out the inference. We present approximation algorithms that provide provably good estimates of edge probabilities. We also present results from an experimental evaluation of our algorithms on several real-world networks.

Trajectory compression algorithms enable efficient transmission, storage, and processing of trajectory data by eliminating redundant information. While a large number of compression algorithms have been developed, there is no comprehensive and convenient benchmarking system for evaluating these algorithms. We will demonstrate TrajMetrix, our system that meets the above need. We will show how TrajMetrix can be used to gain insights into the benefits and drawbacks of various compression algorithms given different compression requirements.

Eliminating interactions among individuals is an important means of blocking contagion spread, e.g., closing schools during an epidemic or shutting down electronic communication channels during social unrest. We study contagion blocking in networked populations by identifying edges to remove from a network, thus blocking contagion transmission pathways. We formulate various problems to minimize contagion spread and show that some are efficiently solvable while others are formally hard. We also compare our hardness results to those from node blocking problems and show interesting differences between the two. Our main problem is not only hard, but also has no approximation guarantee, unless P=NP. Therefore, we devise a heuristic for the problem and compare its performance to state-of-the-art heuristics from the literature. We show, through results of 12 (network, heuristic) combinations on three real social networks, that our method offers considerable improvement in the ability to block contagions in weighted and unweighted networks. We also conduct a parametric study to understand the limitations of our approach.

The massive volumes of trajectory data generated by inexpensive GPS devices have led to difficulties in processing, querying, transmitting and storing such data. To overcome these difficulties, a number of algorithms for compressing trajectory data have been proposed. These algorithms try to reduce the size of trajectory data, while preserving the quality of the information. We present results from a comprehensive empirical evaluation of many compression algorithms including Douglas-Peucker Algorithm, Bellman's Algorithm, STTrace Algorithm and Opening Window Algorithms. Our empirical study uses different types of real-world data such as pedestrian, vehicle and multimodal trajectories. The algorithms are compared using several criteria including execution times and the errors caused by compressing spatio-temporal information, across numerous real-world datasets and various error metrics.

The area of clustering under constraints has recently received much attention in the data mining community. However, most work involves adding constraints to existing algorithms which, although being quite pragmatic, raises several difficulties. Examples of these difficulties include creating intractable constraint satisfaction sub-problems and constrained clustering algorithms that are easily over-constrained so they may not converge or converge to a poor clustering solution. In this paper we show how both instance and cluster-level constraints can be expressed as instances of the 2SAT problem and how multiple calls to a 2SAT solver can be used to construct algorithms that are guaranteed to satisfy all the constraints and converge to a global optimum for a number of intuitive objective functions. Our approach provides two additional advantages. Firstly, it leads to polynomial time algorithms for the k = 2 case for several objective functions. Secondly, one can specify large sets of constraints without fear of over-constraining the problem: if one or more solutions satisfying all constraints exist, our algorithm is guaranteed to find a best such solution. We present experimental results to show that our approach outperforms several popular algorithms particularly for large constraint sets, where these algorithms are over-constrained and fair poorly.

Clustering with constraints is an emerging area of data mining research. However, most work assumes that the constraints are given as one large batch. In this paper we explore the situation where the constraints are incrementally given. In this way the user after seeing a clustering can provide positive and negative feedback via constraints to critique a clustering solution. We consider the problem of efficiently updating a clustering to satisfy the new and old constraints rather than reclustering the entire data set. We show that the problem of incremental clustering under constraints is NP-hard in general, but identify several sufficient conditions which lead to efficiently solvable versions. These translate into a set of rules on the types of constraints thatcan be added and constraint set properties that must be maintained. We demonstrate that this approach is more efficient than re-clustering the entire data set and has several other advantages.

Recent work has looked at extending the k-Means algorithm to incorporate background information in the form of instance level must-link and cannot-link constraints. We introduce two ways of specifying additional background information in the form of δ and ∊ constraints that operate on all instances but which can be interpreted as conjunctions or disjunctions of instance level constraints and hence are easy to implement. We present complexity results for the feasibility of clustering under each type of constraint individually and several types together. A key finding is that determining whether there is a feasible solution satisfying all constraints is, in general, NP-complete. Thus, an iterative algorithm such as k-Means should not try to find a feasible partitioning at each iteration. This motivates our derivation of a new version of the k-Means algorithm that minimizes the constrained vector quantization error but at each iteration does not attempt to satisfy all constraints. Using standard UCI datasets, we find that using constraints improves accuracy as others have reported, but we also show that our algorithm reduces the number of iterations until convergence. Finally, we illustrate these benefits and our new constraint types on a complex real world object identification problem using the infra-red detector on an Aibo robot.

Commercial distributed database systems generally support an optional protocol that provides loose consistency of replicas, allowing replicas to be inconsistent for some time. In such a protocol, each replicated data item is assigned a primary copy site. Typically, a transaction updates only the primary copies of data items, with updates to other copies deferred until after the transaction commits. After a transaction commits, its updates to primary copies are sent transactionally to the other sites containing secondary copies. We investigate the transaction model underlying the above protocol. We show that global serializability in such a system is a property of the placement of primary and secondary copies of replicated data items. We present a polynomial time algorithm to assign primary sites to data items so that the resulting topology ensures serializability.