We address the problem of combining proximity and semantic criteria for detecting co-moving clusters of interest in atomic trajectories. Specifically, we are interested in the motion of atoms from different molecules that at some point form a Hydrogen Bond (HB) and that HB persists over time with additional constraints within clusters. Moreover, it is permissible that an HB within a cluster is disrupted for a brief period. To enable the detection of such phenomena, we introduce the notion of Bond-Aware Relaxed Moving Clusters (BARMC) pattern and present a Naïve algorithm for its detection.

We present HydroFlow, a novel deep generative model for predicting the electricity generation demand of large-scale hydropower stations. HydroFlow uses a latent stochastic recurrent neural network to capture the dependencies in the multivariate time series. It not only utilizes the hidden state of the neural network, but also considers the uncertainty of variables related to natural and social factors. We also introduce an end-to-end approach based on generative flows to approximate the posterior distribution of time series with exact likelihoods. Our model is powerful as adding stochasticity to different factors (e.g., reservoir capacity and water-flow measurements) and thus overcomes the expressiveness limitations of deterministic prediction methods. It also enables trainable latent transformations that can improve the model interpretability. We evaluate HydroFlow on the data collected from the hydropower stations of a large-scale hydropower development company. Experimental results show that our model significantly outperforms the state-of-the-art baseline methods while providing explainable results.

We present a prototype system for planning shopping-related trips in the settings in which an individual may need to purchase a collection of items for which the requests originate from different categories of groups. Specifically, we consider scenarios in which the user is affiliated with a family and other (multiple) social circles - with varying temporal duration (e.g., from a specific party/event, to longer lasting preference-based memberships such as chess playing, reading club, etc.). Each of the groups may have its separate list of items to be purchased at a given time - e.g., grocery for the family; wine and cheese for a social event - however: (1) a particular individual may be a member of multiple (different) groups; (2) a particular group may have different shopping list (for different purposes); and (3) a user may be at different location at times when different requests originate. Our UGASP (User and Group Aware Shopping Planner) system aims at: (1) Generating trajectories for individuals to complete part of the purchase list, based on their group memberships; (2) Calculating new trajectories, based on updates of the purchase assignments to different users.

Network alignment (NA) aims to link common nodes across multiple networks and is an essential task in many graph mining applications. Despite the progress achieved by many recent works, several fundamental limitations have eluded the proper cohesive way of addressing, including matching confusion, lack of the formal treatment of uncertainty, and Point-to-Point (P2P) constraint. This study proposes a novel framework UANA (Uncertainty-Aware Network Alignment) to tackle the limitations of the existing works. By embedding nodes as Gaussian distributions rather than point vectors, UANA enables to capture the uncertainty of a node representation, while being able to discriminate the anchor nodes from the potentially confusing neighbors. We address the P2P matching constraint by introducing an adversarial learning paradigm, which relaxes the exact matching assumption during training with an across-domain generative procedure to reduce the matching errors on testing nodes. In the end, interpretability methods are included to explain the aligning results made by our UANA based on the robust statistics, which enables the explanation of the effect of individual training sample on the NA performance without the need of retraining the model. Extensive experiments conducted on real-world data sets demonstrate that UANA significantly outperforms existing state-of-the-art baselines while providing explainable results.

With the rapid growth of electronic data repositories in diverse application domains, including healthcare, considerable research interest has been developed to solve issues related to extraction of hidden knowledge in these repositories. Electronic health record systems (EHRs) are the fastest growing in terms of size and data diversity. In this work, we focus on mining a high dimensional sparse dataset using nursing care data as an exemplar. To mine a high-dimensional and sparse dataset is a challenging task due to a number of reasons. There are several dimension reduction methods, however, they do not work well with contextual datasets. In our study, we have used association mining as a dimension reduction step and for extracting important features from the dataset. Our results show that association mining can be effectively used for dimension reduction and feature extraction step. Our predictive modeling results show that decision tree models generally have high accuracy and the results are easy to interpret and determine the influence of different variables.

This work addresses the problem of energy efficient management of mobile resource distribution in Wireless Sensor Networks (WSN), subject to Quality of Service (QoS) constraints. Monitored phenomena may require an increased coverage within a particular area and we present novel methodologies for optimizing the "bargaining stage" when deciding how to select the mobile resources to be re-located in response to such events. Our experimental results demonstrate significant energy savings, both in terms of communication overheads and maintenance of the hierarchical routing structures, as well as the quality assurances in terms of the turnaround time.

We address the problem of efficient in-network processing of hybrid spatial queries in Wireless Sensor Networks (WSN), where the data may correspond to different physical phenomena in different regions. We propose space and communication efficient schemes capable of correlating spatial dimension with different physical values. To trade-off (im)precision vs. energy consumption, the proposed schemes combine rank order statistics, regular sampling, and bitmap representation. We present a proof of concept implementation of the proposed methodology and quantify the benefits of our approach through simulations.

Wireless sensor networks (WSN) have emerged as a powerful paradigm for real-time monitoring of physical phenomena. Depending on WSN application, indexing of sensed data is an important issue that deals with in-network organizing of the data and subsequently answering queries, with minimum power consumption. Several approaches have been developed to build data indexing systems in sensor networks. However, most of these systems are suitable to specific types of spatial or temporal queries, and entail significant communication overhead, thus creating unbalanced loads and affecting network lifetime. In this paper we present a novel data indexing system that is based on multiple hierarchical channels of data and physical space abstraction, and is capable of answering spatio-temporal queries. The proposed system also allows approximate queries to be answered quickly and efficiently.

This paper presents a scalable content-based image indexing and retrieval system based on vector wavelet coefficients of color images. Highly decorrelated wavelet coefficient planes are used to acquire a search efficient feature space. The feature space is subsequently indexed using properties of all the images in the database. Therefore, the feature key of an image not only corresponds to the content of the image itself but also to how much the image is different from the other images being stored in the database. The search time linearly depends on the number of images similar to the query image and is independent of the database size. We show that, in a database of 5,000 images, query search takes less than 30 msec on a 266 MHz Pentium II processor, compared to several seconds of retrieval time in the earlier systems proposed in the literature.

We propose a graphical data model for specifying spatio-temporal semantics of video data. The proposed model segments a video clip into subsegments consisting of objects. Each object is detected and recognized, and the relevant information of each object is recorded. The motions of objects are modeled through their relative spatial relationships as time evolves. Based on the semantics provided by this model, a user can create his/her own, object-oriented view of the video database. Using the propositional logic, we describe a methodology for specifying conceptual queries involving spatio-temporal semantics and expressing views for retrieving various video clips. Alternatively, a user can sketch the query, by exemplifying the concept. The proposed methodology can be used to specify spatio-temporal concepts at various levels of information granularity.>