Abstract

Abstract
Given an heterogeneous social network, can we forecast its future? Can we predict who will start using a given hashtag on twitter? Can we leverage side information, such as who retweets or follows whom, to improve our membership forecasts? We present TENSORCAST, a novel method that forecasts time-evolving networks more accurately than current state-of-the-art methods by incorporating multiple data sources in coupled tensors. TENSORCAST is (a) scalable, being linearithmic on the number of connections; (b) effective, achieving over 20% improved precision on top-1000 forecasts of community members; (c) general, being applicable to data sources with different structure. We run our method on multiple real-world networks, including DBLP, epidemiology data, power grid data, and a Twitter temporal network with over 310 million nonzeros, where we predict the evolution of the activity of the use of political hashtags.

Abstract
Habitat mapping is an important task to manage ecosystems. This task becomes most challenging when it comes to marine habitats as it is hard to get good images in underwater conditions and to precisely locate them. In this paper we present a novel technique for performing habitat mapping automating all phases, from data collection to classification, lowering costs and increasing efficiency throughout the process. For mapping habitats in a vast coastal region, we use visible light cameras mounted on autonomous underwater vehicles, capable of collecting and geo-locating all acquired data. The optic images are enhanced using Computer Vision techniques, to help specialists identify the habitats they contain (during training phase). In a later stage, we employ convolutional neural networks to automatically identify habitats in all imagery. Habitats are classified according to the European Nature Information System, an European classification standard for habitats.

Abstract
Finding important nodes is a classic task in network science. Nodes are important depending on the context; e.g., they can be (i) nodes that, when removed, cause the network to collapse or (ii) influential spreaders (e.g., of information, or of diseases). Typically, central nodes are assumed to be important, and numerous network centrality measures have been proposed such as the degree centrality, the betweenness centrality, and the subgraph centrality. However, centrality measures are not tailored to capture one particular kind of important nodes: dominant nodes. We define dominant nodes as nodes that dominate many others and are not dominated by many others. We then propose a general graphlet-based measure of node dominance called graphlet-dominance (GD). We analyze how GD differs from traditional network centrality measures. We also study how certain parameters (namely the importance of dominating versus not being dominated and indirect versus direct dominances) influence GD. Finally, we apply GD to author ranking and verify that GD is superior to PageRank in four of the five citation networks tested. © 2020, Springer Nature Switzerland AG.

Abstract
Counting subgraph occurrences is a hard but very important task in complex network analysis, with applications in concepts such as network motifs or graphlet degree distributions. In this paper we present a novel approach for this task that takes advantage of knowing that a large fraction of subgraph types does not appear at all on real-world networks. We describe a pattern-growth methodology that is able to iteratively build subgraph patterns that do not contain smaller non-occurring subgraphs, significantly pruning the search space. By using the g-trie data structure, we are able to efficiently only count those subgraphs that we are interested in, reducing the total computation time. The obtained experimental results are very promising allowing us to avoid the computation of up to 99.78% of all possible subgraph patterns. This showcases the potential of this approach and paves the way for reaching previously unattainable subgraph sizes. © Springer Nature Switzerland AG 2019.

Abstract
Scientific impact is commonly associated with the number of citations received. However, an author can easily boost his own citation count by (i) publishing articles that cite his own previous work (self-citations), (ii) having co-authors citing his work (co-author citations), or (iii) exchanging citations with authors from other research groups (reciprocated citations). Even though these friendly citations inflate an author's perceived scientific impact, author ranking algorithms do not normally address them. They, at most, remove self-citations. Here we present Friends-Only Citations AnalySer (FOCAS), a method that identifies friendly citations and reduces their negative effect in author ranking algorithms. FOCAS combines the author citation network with the co-authorship network in order to measure author proximity and penalises citations between friendly authors. FOCAS is general and can be regarded as an independent module applied while running (any) PageRank-like author ranking algorithm. FOCAS can be tuned to use three different criteria, namely authors' distance, citation frequency, and citation recency, or combinations of these. We evaluate and compare FOCAS against eight state-of-the-art author ranking algorithms. We compare their rankings with a ground-truth of best paper awards. We test our hypothesis on a citation and co-authorship network comprised of seven Information Retrieval top-conferences. We observed that FOCAS improved author rankings by 25% on average and, in one case, leads to a gain of 46%.