Search

In this paper we present a novel method, $\textit{Knowledge Persistence}$ ($\mathcal{KP}$), for faster evaluation of Knowledge Graph (KG) completion approaches. Current ranking-based evaluation is quadratic in the size of the KG, leading to long evaluation times and consequently a high carbon footprint. $\mathcal{KP}$ addresses this by representing the topology of the KG completion methods through the lens of topological data analysis, concretely using persistent homology. The characteristics of persistent homology allow $\mathcal{KP}$ to evaluate the quality of the KG completion looking only at a fraction of the data. Experimental results on standard datasets show that the proposed metric is highly correlated with ranking metrics (Hits@N, MR, MRR). Performance evaluation shows that $\mathcal{KP}$ is computationally efficient: In some cases, the evaluation time (validation+test) of a KG completion method has been reduced from 18 hours (using Hits@10) to 27 seconds (using $\mathcal{KP}$), and on average (across methods & data) reduces the evaluation time (validation+test) by $\approx$ $\textbf{99.96}\%$., To appear in proceedings of The Web Conference 2023 (WWW'23)

Graph representation learning resurges as a trending research subject owing to the widespread use of deep learning for Euclidean data, which inspire various creative designs of neural networks in the non-Euclidean domain, particularly graphs. With the success of these graph neural networks (GNN) in the static setting, we approach further practical scenarios where the graph dynamically evolves. Existing approaches typically resort to node embeddings and use a recurrent neural network (RNN, broadly speaking) to regulate the embeddings and learn the temporal dynamics. These methods require the knowledge of a node in the full time span (including both training and testing) and are less applicable to the frequent change of the node set. In some extreme scenarios, the node sets at different time steps may completely differ. To resolve this challenge, we propose EvolveGCN, which adapts the graph convolutional network (GCN) model along the temporal dimension without resorting to node embeddings. The proposed approach captures the dynamism of the graph sequence through using an RNN to evolve the GCN parameters. Two architectures are considered for the parameter evolution. We evaluate the proposed approach on tasks including link prediction, edge classification, and node classification. The experimental results indicate a generally higher performance of EvolveGCN compared with related approaches. The code is available at \url{https://github.com/IBM/EvolveGCN}., Comment: AAAI 2020. The code is available at https://github.com/IBM/EvolveGCN

Recent developments in machine learning have enabled accurate predictions of the dynamics of slow structural relaxation in glass-forming systems. However, existing machine-learning models for these tasks are mostly designed such that they learn a single dynamic quantity and relate it to the structural features of glassy liquids. In this study, we propose a graph neural network model, ``BOnd TArgeting Network (BOTAN)'', that learns relative motion between neighboring pairs of particles, in addition to the self-motion of particles. By relating the structural features to these two different dynamical variables, the model autonomously acquires the ability to discern how different dynamical processes, strain fluctuations and particle rearrangements, affect the self-motion of particles undergoing slow relaxation, and thus can predict with high precision how slow structural relaxation develops in space and time., Comment: 11 pages, 10 figures, a typo fixed in the title; accepted in J. Chem. Phys

We propose Distributed Neighbor Expansion (Distributed NE), a parallel and distributed graph partitioning method that can scale to trillion-edge graphs while providing high partitioning quality. Distributed NE is based on a new heuristic, called parallel expansion, where each partition is constructed in parallel by greedily expanding its edge set from a single vertex in such a way that the increase of the vertex cuts becomes local minimal. We theoretically prove that the proposed method has the upper bound in the partitioning quality. The empirical evaluation with various graphs shows that the proposed method produces higher-quality partitions than the state-of-the-art distributed graph partitioning algorithms. The performance evaluation shows that the space efficiency of the proposed method is an order-of-magnitude better than the existing algorithms, keeping its time efficiency comparable. As a result, Distributed NE can partition a trillion-edge graph using only 256 machines within 70 minutes., Comment: VLDB 2020, Code in http://www.masahanai.jp/DistributedNE/

Using computer simulation to analyze large-scale discrete event systems requires repeated executions with various scenarios or parameters. Such repeated executions can induce significant redundancy in event processing when the modification from a prior scenario to a new scenario is relatively minor, and when the altered scenario influences only a small part of the simulation. For example, in a city-scale traffic simulation, an altered scenario of blocking one junction may only affect a small part of the city for considerable length of time. However, traditional simulation approaches would still repeat the simulation for the whole city even when the changes are minor. In this article, we propose a new redundancy reduction technique for large-scale discrete event simulations, called exact-differential simulation , which simulates only the altered portions of scenarios and their influences in repeated executions while still achieving the same results as the re-execution of entire simulations. This article presents the main concepts of the exact-differential simulation, the design of its algorithm, and an approach to build an exact-differential simulation middleware that supports multiple applications of discrete event simulation. We also evaluate our approach by using two case studies, PHOLD benchmark and a traffic simulation of Tokyo.

The COVID-19 pandemic is changing the world in unprecedented and unpredictable ways. Human mobility, being the greatest facilitator for the spread of the virus, is at the epicenter of this change. In order to study mobility under COVID-19, to evaluate the efficiency of mobility restriction policies, and to facilitate a better response to future crisis, we need to understand all possible mobility data sources at our disposal. Our work studies private mobility sources, gathered from mobile-phones and released by large technological companies. These data are of special interest because, unlike most public sources, it is focused on individuals rather than on transportation means. Furthermore, the sample of society they cover is large and representative. On the other hand, these data are not directly accessible for anonymity reasons. Thus, properly interpreting its patterns demands caution. Aware of that, we explore the behavior and inter-relations of private sources of mobility data in the context of Spain. This country represents a good experimental setting due to both its large and fast pandemic peak and its implementation of a sustained, generalized lockdown. Our work illustrates how a direct and naive comparison between sources can be misleading, as certain days (e.g., Sundays) exhibit a directly adverse behavior. After understanding their particularities, we find them to be partially correlated and, what is more important, complementary under a proper interpretation. Finally, we confirm that mobile-data can be used to evaluate the efficiency of implemented policies, detect changes in mobility trends, and provide insights into what new normality means in Spain. Part of this research has received funding from the European Union Horizon 2020 Programme under the SoBigData++ Project, grant agreement num. 871,042. Martí Català received funding from La Caixa Foundation (ID 100010434), under agreement LCF/PR/GN17/50300003; Martí Català received funding from Ministerio de Ciencia, Innovación y Universidades and FEDER, with the project PGC2018-095456-B-I00.E. A-L EAL thanks support from the Spanish Ministry of Science Innovation and Universities, (SAF2017-88019-C3-3R).

We present distributed algorithms for training dynamic Graph Neural Networks (GNN) on large scale graphs spanning multi-node, multi-GPU systems. To the best of our knowledge, this is the first scaling study on dynamic GNN. We devise mechanisms for reducing the GPU memory usage and identify two execution time bottlenecks: CPU-GPU data transfer; and communication volume. Exploiting properties of dynamic graphs, we design a graph difference-based strategy to significantly reduce the transfer time. We develop a simple, but effective data distribution technique under which the communication volume remains fixed and linear in the input size, for any number of GPUs. Our experiments using billion-size graphs on a system of 128 GPUs shows that: (i) the distribution scheme achieves up to 30x speedup on 128 GPUs; (ii) the graph-difference technique reduces the transfer time by a factor of up to 4.1x and the overall execution time by up to 40%, Comment: Conference version to appear in the proceedings of SC'21

Inspired by the generation power of generative adversarial networks (GANs) in image domains, we introduce a novel hierarchical architecture for learning characteristic topological features from a single arbitrary input graph via GANs. The hierarchical architecture consisting of multiple GANs preserves both local and global topological features and automatically partitions the input graph into representative “stages” for feature learning. The stages facilitate reconstruction and can be used as indicators of the importance of the associated topological structures. The experiments show that our method produces subgraphs retaining a wide range of topological features, even in early reconstruction stages (unlike a single GAN, which cannot easily identify such features, let alone reconstruct the original graph). This paper is the firstline research on combining the use of GANs and graph topological analysis.

Vector-space word representations obtained from neural network models have been shown to enable semantic operations based on vector arithmetic. In this paper, we explore the existence of similar information on vector representations of images. For that purpose we define a methodology to obtain large, sparse vector representations of image classes, and generate vectors through the state-of-the-art deep learning architecture GoogLeNet for 20K images obtained from ImageNet. We first evaluate the resultant vector-space semantics through its correlation with WordNet distances, and find vector distances to be strongly correlated with linguistic semantics. We then explore the location of images within the vector space, finding elements close in WordNet to be clustered together, regardless of significant visual variances (e.g. 118 dog types). More surprisingly, we find that the space unsupervisedly separates complex classes without prior knowledge (e.g. living things). Afterwards, we consider vector arithmetics. Although we are unable to obtain meaningful results on this regard, we discuss the various problem we encountered, and how we consider to solve them. Finally, we discuss the impact of our research for cognitive systems, focusing on the role of the architecture being used., Comment: 14 pages, 5 figures, accepted at Cognitive Systems Research

The outbreak of coronavirus disease 2019 (COVID-19) recently has affected human life to a great extent. Besides direct physical and economic threats, the pandemic also indirectly impact people’s mental health conditions, which can be overwhelming but difficult to measure. The problem may come from various reasons such as unemployment status, stay-at-home policy, fear for the virus, and so forth. In this work, we focus on applying natural language processing (NLP) techniques to analyze tweets in terms of mental health. We trained deep models that classify each tweet into the following emotions: anger, anticipation, disgust, fear, joy, sadness, surprise and trust. We build the EmoCT (Emotion-Covid19-Tweet) dataset for the training purpose by manually labeling 1,000 English tweets. Furthermore, we propose an approach to find out the reasons that are causing sadness and fear, and study the emotion trend in both keyword and topic level.

As COVID-19 transmissions spread worldwide, governments have announced and enforced travel restrictions to prevent further infections. Such restrictions have a direct effect on the volume of international flights among these countries, resulting in extensive social and economic costs. To better understand the situation in a quantitative manner, we analyzed the OpenSky Network data to clarify flight patterns and flight densities around the world. Then we observed relationships between flight numbers with new infection cases and the economy (the unemployment rate) in Barcelona. We found that the number of daily flights gradually decreased and then suddenly dropped 64% during the second half of March in 2020 after the United States and Europe enacted travel restrictions. We also observed a 51% decrease in the global flight network density decreased during this period. Regarding new COVID-19 cases, the United States had an unexpected surge regardless of travel restrictions. Finally, the layoffs for temporary workers in the tourism and airplane business increased by 4.3 fold in the weeks following Spain’s decision to close its borders.

There are many large-scale graphs in real world such as Web graphs and social graphs. The interest in large-scale graph analysis is growing in recent years. Breadth-First Search (BFS) is one of the most fundamental graph algorithms used as a component of many graph algorithms. Our new method for distributed parallel BFS can compute BFS for one trillion vertices graph within half a second, using large supercomputers such as the K-Computer. By the use of our proposed algorithm, the K-Computer was ranked 1st in Graph500 using all the 82,944 nodes available on June and November 2015 and June 2016 38,621.4 GTEPS. Based on the hybrid BFS algorithm by Beamer (Proceedings of the 2013 IEEE 27th International Symposium on Parallel and Distributed Processing Workshops and PhD Forum, IPDPSW ’13, IEEE Computer Society, Washington, 2013), we devise sets of optimizations for scaling to extreme number of nodes, including a new efficient graph data structure and several optimization techniques such as vertex reordering and load balancing. Our performance evaluation on K-Computer shows that our new BFS is 3.19 times faster on 30,720 nodes than the base version using the previously known best techniques.

Graph pattern matching algorithms to handle million-scale dynamic graphs are widely used in many applications such as social network analytics and suspicious transaction detections from financial networks. On the other hand, the computation complexity of many graph pattern matching algorithms is expensive, and it is not affordable to extract patterns from million-scale graphs. Moreover, most real-world networks are time-evolving, updating their structures continuously, which makes it harder to update and output newly matched patterns in real time. Many incremental graph pattern matching algorithms which reduce the number of updates have been proposed to handle such dynamic graphs. However, it is still challenging to recompute vertices in the incremental graph pattern matching algorithms in a single process, and that prevents the real-time analysis. We propose an incremental graph pattern matching algorithm to deal with time-evolving graph data and also propose an adaptive optimization system based on reinforcement learning to recompute vertices in the incremental process more efficiently. Then we discuss the qualitative efficiency of our system with several types of data graphs and pattern graphs. We evaluate the performance using million-scale attributed and time-evolving social graphs. Our incremental algorithm is up to 10.1 times faster than an existing graph pattern matching and 1.95 times faster with the adaptive systems in a computation node than naive incremental processing., Comment: 10 pages and 11 figures

Deep neural networks are representation learning techniques. During training, a deep net is capable of generating a descriptive language of unprecedented size and detail in machine learning. Extracting the descriptive language coded within a trained CNN model (in the case of image data), and reusing it for other purposes is a field of interest, as it provides access to the visual descriptors previously learnt by the CNN after processing millions of images, without requiring an expensive training phase. Contributions to this field (commonly known as feature representation transfer or transfer learning) have been purely empirical so far, extracting all CNN features from a single layer close to the output and testing their performance by feeding them to a classifier. This approach has provided consistent results, although its relevance is limited to classification tasks. In a completely different approach, in this paper we statistically measure the discriminative power of every single feature found within a deep CNN, when used for characterizing every class of 11 datasets. We seek to provide new insights into the behavior of CNN features, particularly the ones from convolutional layers, as this can be relevant for their application to knowledge representation and reasoning. Our results confirm that low and middle level features may behave differently to high level features, but only under certain conditions. We find that all CNN features can be used for knowledge representation purposes both by their presence or by their absence, doubling the information a single CNN feature may provide. We also study how much noise these features may include, and propose a thresholding approach to discard most of it. All these insights have a direct application to the generation of CNN embedding spaces., Published in the Journal of Artificial Intelligence Research (JAIR), Special Track on Deep Learning, Knowledge Representation, and Reasoning

Community detections for large-scale real world networks have been more popular in social analytics. In particular, dynamically growing network analyses become important to find long-term trends and detect anomalies. In order to analyze such networks, we need to obtain many snapshots and apply same analytic methods to them. However, it is inefficient to extract communities from these whole newly generated networks with little differences every time, and then it is impossible to follow the network growths in the real time. We proposed an incremental community detection algorithm for high-volume graph streams. It is based on the top of a well-known batch-oriented algorithm named DEMON[1]. We also evaluated performance and precisions of our proposed incremental algorithm with real-world big networks with up to 410,236 vertices and 2,439,437 edges and computed in less than one second to detect communities in an incremental fashion - which achieves up to 107 times faster than the original algorithm without sacrificing accuracies., Comment: 8 pages, 7 figures and 3 pseudo codes, 2016 IEEE International Conference on Big Data (Big Data)

Finding communities in multilayer networks is a vital step in understanding the structure and dynamics of these layers, where each layer represents a particular type of relationship between nodes in the natural world. However, most community discovery methods for multilayer networks may ignore the interplay between layers or the unique topological structure in a layer. Moreover, most of them can only detect non-overlapping communities. In this paper, we propose a new community discovery method for multilayer networks, which leverages the interplay between layers and the unique topology in a layer to reveal overlapping communities. Through a comprehensive analysis of edge behaviors within and across layers, we first calculate the similarities for edges from the same layer and the cross layers. Then, by leveraging these similarities, we can construct a dendrogram for the multilayer networks that takes both the unique topological structure and the important interplay into consideration. Finally, by introducing a new community density metric for multilayer networks, we can cut the dendrogram to get the overlapping communities for these layers. By applying our method on both synthetic and real-world datasets, we demonstrate that our method has an accurate performance in discovering overlapping communities in multilayer networks.

Multilayer network analysis has become a vital tool for understanding different relationships and their interactions in a complex system, where each layer in a multilayer network depicts the topological structure of a group of nodes corresponding to a particular relationship. The interactions among different layers imply how the interplay of different relations on the topology of each layer. For a single-layer network, network embedding methods have been proposed to project the nodes in a network into a continuous vector space with a relatively small number of dimensions, where the space embeds the social representations among nodes. These algorithms have been proved to have a better performance on a variety of regular graph analysis tasks, such as link prediction, or multi-label classification. In this paper, by extending a standard graph mining into multilayer network, we have proposed three methods ("network aggregation," "results aggregation" and "layer co-analysis") to project a multilayer network into a continuous vector space. On one hand, without leveraging interactions among layers, "network aggregation" and "results aggregation" apply the standard network embedding method on the merged graph or each layer to find a vector space for multilayer network. On the other hand, in order to consider the influence of interactions among layers, "layer co-analysis" expands any single-layer network embedding method to a multilayer network. By introducing the link transition probability based on information distance, this method not only uses the first and second order random walk to traverse on a layer, but also has the ability to traverse between layers by leveraging interactions. From the evaluation, we have proved that comparing with regular link prediction methods, "layer co-analysis" achieved the best performance on most of the datasets, while "network aggregation" and "results aggregation" also have better performance than regular link prediction methods.

The authors have inadvertently left out three coauthors from the above paper [1] . The names of the three authors are Hal Cooper, Min-Hwan Oh, and Sailung Yeung.

Our claim is that so-called "Big Data" will evolve into a new era with proliferation of data from multiple sources such as massive numbers of sensors whose resolution is increasing expo- nentially, high-resolution simulations generating huge data results, as well as evolution of social infrastructures that allow for "opening up of data silos", i.e., data sources being abundant across the world instead of being confined within an institution, much as how scientific data are being handled in the modern era as a common asset openly accessible within and across disciplines. Such a situation would create the need for not only petabytes to zetabytes of capacity and beyond, but also for extreme scale computing power. Our new project, sponsored under the Japanese JST- CREST program is called "Extreme Big Data", and aims to achieve the convergence of extreme supercomputing and big data in order to cope with such explosion of data. The project consists of six teams, three of which deals with defining future EBD convergent SW/HW architecture and system, and the other three the EBD co-design applications that represent di?erent facets of big data, in metagenomics, social simulation, and climate simulation with real-time data assimilation. Although the project is still early in its lifetime, started in Oct. 2013, we have already achieved several notable results, including becoming world #1 on the Green Graph 500, a benchmark to measure the power e?ciency of graph processing that appear in typical big data scenarios.

Online graph database service providers have started migrating their operations to public clouds due to the increasing demand for low-cost, ubiquitous graph data storage and analysis. However, there is little support available for benchmarking graph database systems in cloud environments. We describe XGDBench which is a graph database benchmarking platform for cloud computing systems. XGDBench has been designed with the aim of creating an extensible platform for graph database benchmarking which makes it suitable for benchmarking future HPC systems. We extend the Yahoo! Cloud Serving Benchmark (YCSB) to the area of graph database benchmarking by creation of XGDBench. The benchmarking platform is written in X10 which is a PGAS language intended for programming future HPC systems. We describe the architecture of the XGDBench and explain how it differs from the current state-of-the-art. We conduct performance evaluation of five famous graph data stores AllegroGraph, Fuseki, Neo4j, OrientDB, and Titan using XGDBench on Tsubame 2.0 HPC cloud environment.

Distributed data stream processing is a data analysis paradigm where massive amounts of data produced by various sources are analyzed online within real-time constraints. Execution performance of a stream program/query executed on such middleware is largely dependent on the ability of the programmer to fine tune the program to match the topology of the stream processing system. However, manual fine tuning of a stream program is a very difficult, error prone process that demands huge amounts of programmer time and expertise which are expensive to obtain. We describe an automated process for stream program performance optimization that uses semantic preserving automatic code transformation to improve stream processing job performance. We first identify the structure of the input program and represent the program structure in a Directed Acyclic Graph. We transform the graph using the concepts of Tri-OP Transformation and Bi-Op Transformation. The resulting sample program space is pruned using both empirical as well as profiling information to obtain a ranked list of sample programs which have higher performance compared to their parent program. We successfully implemented this methodology on a prototype stream program performance optimization mechanism called Hirundo. The mechanism has been developed for optimizing SPADE programs which run on System S stream processing run-time. Using five real world applications (called VWAP, CDR, Twitter, Apnoea, and Bargain) we show the effectiveness of our approach. Hirundo was able to identify a 31.1 times higher performance version of the CDR application within seven minutes time on a cluster of 4 nodes.

We introduce a community detection algorithm (Fluid Communities) based on the idea of fluids interacting in an environment, expanding and contracting as a result of that interaction. Fluid Communities is based on the propagation methodology, which represents the state-of-the-art in terms of computational cost and scalability. While being highly efficient, Fluid Communities is able to find communities in synthetic graphs with an accuracy close to the current best alternatives. Additionally, Fluid Communities is the first propagation-based algorithm capable of identifying a variable number of communities in network. To illustrate the relevance of the algorithm, we evaluate the diversity of the communities found by Fluid Communities, and find them to be significantly different from the ones found by alternative methods., Accepted at the 6th International Conference on Complex Networks and Their Applications

Informatics for Integrating Biology and the Bedside (i2b2) is a database system to facilitate sharing and reuse of clinical patients' data collected in individual hospitals. The i2b2 provides an ontology based object-oriented database system with highly simple and flexible database schema which enables us to integrate clinical patients' data from different laboratories and different hospitals. 392 patients' data including carcinoma and non-carcinoma specimens from cancer patients are transported from the Integrated Clinical Omics Database (iCOD) to the i2b2 database for a feasibility study to check applicability of i2b2 ontology and database schema on Japanese clinical patients' data. No modification is required for the i2b2 data model to deal with Japanese characters. Some modification of ontology is required to integrate biomedical information extracted from the cancer patients' data. We believe that the i2b2 system will be practical infrastructure to integrate Japanese clinical databases if appropriate disease ontology for Japanese patients is provided.

To analyze and predict a behavior of large-scale traffics with what-if simulation, it needs to repeat many times with various patterns of what-if scenarios. In this paper, we propose new techniques to efficiently repeat what-if simulation tasks with exact-differential simulation. The paper consists of two main efforts: what-if scenario filtering and exact-differential cloning. The what-if scenario filtering enables to pick up meaningful what-if scenarios and reduces the number of what-if scenarios, which directly decreases total execution time of repeating. The exact-differential cloning enables to execute exact-differential simulation tasks in parallel to improve its total execution time. In our preliminary evaluation in Tokyo bay area's traffic simulation, we show potential of our proposals by estimating how the what-if scenarios filtering reduces the number of meaningless scenarios and also by estimating a performance improvement from our previous works with the exact-differential cloning.

Analyzing large-scale traffics by simulation needs repeating execution many times with various patterns of scenarios or parameters. Such repeating execution brings about big redundancy because the change from a prior scenario to a later scenario is very minor in most cases, for example, blocking only one of roads or changing the speed limit of several roads. In this paper, we propose a new redundancy reduction technique, called exact-differential simulation, which enables to simulate only changing scenarios in later execution while keeping exactly same results as in the case of whole simulation. The paper consists of two main efforts: (i) a key idea and algorithm of the exact-differential simulation, (ii) a method to build large-scale traffic simulation on the top of the exact-differential simulation. In experiments of Tokyo traffic simulation, the exact-differential simulation shows 7.26 times as much elapsed time improvement in average and 2.26 times improvement even in the worst case as the whole simulation.

We present a scalable parallel solver for numerical constraint satisfaction problems (NCSPs). Our parallelization scheme consists of homogeneous worker solvers, each of which runs on an available core and communicates with others via the global load balancing (GLB) method. The parallel solver is implemented with X10 that provides an implementation of GLB as a library. In experiments, several NCSPs from the literature were solved and attained up to 516-fold speedup using 600 cores of the TSUBAME2.5 supercomputer., To be presented at X10'15 Workshop

GridRPC, which is an RPC mechanism tailored for the Grid, is an attractive programming model for Grid computing. This paper reports on the design and implementation of a GridRPC programming system called Ninf-G. Ninf-G is a reference implementation of the GridRPC API which has been proposed for standardization at the Global Grid Forum. In this paper, we describe the design, implementations and typical usage of Ninf-G. A preliminary performance evaluation in both WAN and LAN environments is also reported. Implemented on top of the Globus Toolkit, Ninf-G provides a simple and easy programming interface based on standard Grid protocols and the API for Grid Computing. The overhead of remote procedure calls in Ninf-G is acceptable in both WAN and LAN environments.

Using the Cloud for large-scale distributed simulations, such as agent-based traffic simulations, sounds like a good idea, as it is possible to provision and release easily processing nodes (e.g., Virtual machines) in the Cloud. However, the question is complex as it involves users' objectives, such as, time to process the simulation and cost of the simulation, and because the workload evolves in distributed simulations, in each node and the whole system, and this impact the resource provisioning plans. This paper proposes two main contributions: (i) a method for efficient utilization of computational resources for distributed agent-based simulations, providing a mechanism that adapts the resource provisioning to users' objectives and workload evolution, and (ii) a staged asynchronous migration technique to limit the migration overhead when the number of workers change. Our preliminary experimental results on a 24 hour scenario of traffic in the city of Tokyo show that our system outperforms a static provisioning by 12% in average and 23% during periods when workload changes a lot.

Large scale distributed agent-based simulations run on several computing units e.g., virtual machines in the Cloud, nodes in a supercomputer. Classically, these systems try to re-load-balance the nodes as overloaded nodes slow down the process.However another challenge in large scale distributed simulations is that the overall load evolves. In this paper we leverage on commodity computing to adapt resource provisioning number of computing units to the load during the execution of the simulation. We also propose an asynchronous migration mechanism that migrate workload between computing nodes efficiently when nodes wait for synchronisation barriers to happen. We validate our implementation on a scenario simulating one day of vehicular traffic in Tokyo, running on 2 to 8 machines depending on the demand. Our evaluation shows a 26% reduction in data migration time compared to a naive migration approach between computing units.

We present a parallel solver for numerical constraint satisfaction problems (NCSPs) that can scale on a number of cores. Our proposed method runs worker solvers on the available cores and simultaneously the workers cooperate for the search space distribution and balancing. In the experiments, we attained up to 119-fold speedup using 256 cores of a parallel computer.

XML (Extensible Markup Language) processing can incur significant runtime overhead in XML-based infrastructural middleware such as Web service application servers. This paper proposes a novel mechanism for efficiently processing similar XML documents. Given a new XML document as a byte sequence, the XML parser proposed in this paper normally avoids syntactic analysis but simply matches the document with previously processed ones, reusing those results. Our parser is adaptive since it partially parses and then remembers XML document fragments that it has not met before. Moreover, it processes safely since its partial parsing correctly checks the well-formedness of documents. Our implementation of the proposed parser complies with the JSR 63 standard of the Java API for XML Processing (JAXP) 1.1 specification. We evaluated Deltarser performance with messages using Google Web services. Comparing to Piccolo (and Apache Xerces), it effectively parses 35% (106%) faster in a server-side use-case scenario, and 73% (126%) faster in a client-side use-case scenario.

GridSpeed is a grid portal hosting server that automatically generates and publishes a customized Web interface to the grid for applications, with minimal effort required from the user. Users need only to specify information regarding their application using simple GridSpeed Web forms. With GridSpeed, users need not make any modifications to their applications nor write any glue code to publish the application on the Web, not requiring any knowledge of Perl, JSP (Java server pages) or Java servlets. Moreover, the portal generated by GridSpeed provides an application frontend as well as a set of fundamental portal services such as an information service, monitoring service, data management, single sign-on, and so forth. GridSpeed publishes a set of portals as grid services themselves generated by the system that is shamble, searchable, and accessible from others interested in using the application. This feature facilitates the reuse of application portals for specific application domains, as well as increases the number of available grid applications accessible on the Web. This work describes an overview and architecture of the GridSpeed system, and evaluates the system for two real-world scientific applications: BLAST and MCell.

Resilient transportation systems enable quick evacuation, rescue, distribution of relief supplies, and other activities for reducing the impact of natural disasters and for accelerating the recovery from them. The resilience of a transportation system largely relies on the decisions made during a natural disaster. We developed an agent-based traffic simulator for predicting the results of potential actions taken with respect to the transportation system to quickly make appropriate decisions. For realistic simulation, we govern the behavior of individual drivers of vehicles with foundational principles learned from probe-car data. For example, we used the probe-car data to estimate the personality of individual drivers of vehicles in selecting their routes, taking into account various metrics of routes such as travel time, travel distance, and the number of turns. This behavioral model, which was constructed from actual data, constitutes a special feature of our simulator. We built this simulator using the X10 language, which enables massively parallel execution for simulating traffic in a large metropolitan area. We report the use cases of the simulator in three major cities in the context of disaster recovery and resilient transportation.

© 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes,creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. Features extracted through transfer learning can be used to exploit deep learning representations in contexts where there are very few training samples, where there are limited computational resources, or when the tuning of hyper-parameters needed for training deep neural networks is unfeasible. In this paper we propose a novel feature extraction embedding called full-network embedding. This embedding is based on two main points. First, the use of all layers in the network, integrating activations from different levels of information and from different types of layers (i.e., convolutional and fully connected). Second, the contextualisation and leverage of information based on a novel three-valued discretisation method. The former provides extra information useful to extend the characterisation of data, while the later reduces noise and regularises the embedding space. Significantly, this also reduces the computational cost of processing the resultant representations. The proposed method is shown to outperform single layer embeddings on several image classification tasks, while also being more robust to the choice of the pre-trained model used as transfer source. This work is partially supported by the Joint Study Agreement no. W156463 under the IBM/BSC Deep Learning Center agreement, by the Spanish Government through Programa Severo Ochoa (SEV-2015-0493), by the Spanish Ministry of Science and Technology through TIN2015-65316-P project and by the Generalitat de Catalunya (contracts 2014-SGR-1051), and by the Core Research for Evolutional Science and Technology (CREST) program of Japan Science and Technology Agency (JST).