Search

Science gateways have gained increasing attention in the last years from diverse communities. Science gateways are software solutions that bring out the integration of reusable data and specialized techniques via Web servers while hiding the complexity of the underlying high-performance computing resources. Several projects and initiatives have been started worldwide to develop frameworks that support the broad range of key scientific domains. Biological sciences are undergoing a revolution since novel technologies, such as next-generation sequencing, allow data generation in exascale dimensions. Bioinformatics covers a wide range of important applications in health, diversity, and life sciences with the understanding of the high-performance computing culture to accelerate the transition of computational simulations of biological systems at all scales. The article introduces the BioinfoPortal gateway, its architecture, functionalities, and the integration to the CSGrid middleware used to manage the high-performance computing environment of the Brazilian National High-Performance Computing System, SINAPAD, including the Santos Dumont supercomputer. We present a discussion about the challenges of integrating BioinfoPortal and CSGrid framework, which considers the general process of the installation, configuration, and deployment. Finally, we present the findings of the performance analysis of high-performance computing applications, presenting how machine learning was applied to optimize the functionality of BioinfoPortal based on recommending predictive models for the efficient allocation of resources obtained over 75% of performance efficiency.

Consider a set of black-box models – each of them independently trained on a different dataset – answering the same predictive spatio-temporal query. Being built in isolation, each model traverses its own life-cycle until it is deployed to production, learning data patterns from different datasets and facing independent hyper-parameter tuning. In order to answer the query, the set of black-box predictors has to be ensembled and allocated to the spatio-temporal query region. However, computing an optimal ensemble is a complex task that involves selecting the appropriate models and defining an effective allocation strategy that maps the models to the query region. In this paper we present DJEnsemble, a cost-based strategy for the automatic selection and allocation of a disjoint ensemble of black-box predictors to answer predictive spatio-temporal queries. We conduct a set of extensive experiments that evaluate DJEnsemble and highlight its efficiency, selecting model ensembles that are almost as efficient as the optimal solution. When compared against the traditional ensemble approach, DJEnsemble achieves up to 4X improvement in execution time and almost 9X improvement in prediction accuracy.

Astronomy is entering an era of mega-data that will render conventional research methods as well as data and visual analytics tools ineffective. The Square Kilometre Array (SKA) drives one of the most significant big data challenges of the next decades. South Africa, China and India are partners in the global SKA collaboration and host recently completed, next generation radio astronomy facilities. South Africa, Brazil, China and India are involved in the Large Synoptic Survey Telescope (LSST), which represents a complementary mega-data challenge, vastly increasing the current data volume of optical surveys, and providing critical multi-wavelength data set for SKA analytics. Russian researchers are also engaged in radio astronomy and multi-wavelength, multi-messenger projects driving increasing volumes of observational data. This project brings together teams leading programs in data innovation in each partner country to collaborate on the development of new technologies and systems to meet the big data challenge of SKA pathfinder facilities and the multi-wavelength projects that are critical to the advance of astronomy. In so doing we will prototype and demonstrate scalable big data technologies for the new big data era, and establish a BRICS multinational federated data intensive cloud network for collaborative programs in data intensive astronomy.

International audience; We consider big spatial data, which is typically produced in scientific areas such as geological or seismic interpretation. The spatial data can be produced by observation (e.g. using sensors or soil instruments) or numerical simulation programs and correspond to points that represent a 3D soil cube area. However, errors in signal processing and modeling create some uncertainty, and thus a lack of accuracy in identifying geological or seismic phenomenons. Such uncertainty must be carefully analyzed. To analyze uncertainty, the main solution is to compute a Probability Density Function (PDF) of each point in the spatial cube area. However, computing PDFs on big spatial data can be very time consuming (from several hours to even months on a computer cluster). In this paper, we propose a new solution to efficiently compute such PDFs in parallel using Spark, with three methods: data grouping, machine learning prediction and sampling. We evaluate our solution by extensive experiments on different computer clusters using big data ranging from hundreds of GB to several TB. The experimental results show that our solution scales up very well and can reduce the execution time by a factor of 33 (in the order of seconds or minutes) compared with a baseline method.

Applying machine learning models to meteorological data brings many opportunities to the Geosciences field, such as predicting future weather conditions more accurately. In recent years, modeling meteorological data with deep neural networks has become a relevant area of investigation. These works apply either recurrent neural networks (RNN) or some hybrid approach mixing RNN and convolutional neural networks (CNN). In this work, we propose STConvS2S (Spatiotemporal Convolutional Sequence to Sequence Network), a deep learning architecture built for learning both spatial and temporal data dependencies using only convolutional layers. Our proposed architecture resolves two limitations of convolutional networks to predict sequences using historical data: (1) they violate the temporal order during the learning process and (2) they require the lengths of the input and output sequences to be equal. Computational experiments using air temperature and rainfall data from South America show that our architecture captures spatiotemporal context and that it outperforms or matches the results of state-of-the-art architectures for forecasting tasks. In particular, one of the variants of our proposed architecture is 23% better at predicting future sequences and five times faster at training than the RNN-based model used as a baseline., Accepted manuscript. Submitted to Neurocomputing, Elsevier

Deep learning models expect a reasonable amount of training instances to improve prediction quality. Moreover, in classification problems, the occurrence of an unbalanced distribution may lead to a biased model. In this paper, we investigate the problem of species classification from plant images, where some species have very few image samples. We explore reduced versions of imagenet Neural Network winners architecture to filter the space of candidate matches, under a target accuracy level. We show through experimental results using real unbalanced plant image datasets that our approach can lead to classifications within the 5 best positions with high probability.

A 62-year-old previously cognitively intact man presented to the emergency department with acute impairment of anterograde memory (ability to form new memories).1 Head CT did not show acute changes; however, amnesia persisted. MRI disclosed an acute infarct in the left posterior hippocampus and a nonacute lacunar stroke in the right anterior thalamus (figure). Memory testing showed severe deficits in verbal and visual memory, persisting after 2 months. Bilateral lesions located anywhere in the Papez circuit (hippocampal formations, mammillary bodies, anterior thalamic nuclei, cingulate and parahippocampal cortices, and their connections via fornix and mammillothalamic tracts) may be implicated in marked, persistent anterograde memory deficits.1,2 The hippocampus receives its blood supply from collateral branches of the posterior cerebral artery and the anterior choroidal artery.3 The anterior thalamus is supplied by the tuberothalamic artery.4

Science gateways bring out the possibility of reproducible science as they are integrated into reusable techniques, data and workflow management systems, security mechanisms, and high performance computing (HPC). We introduce BioinfoPortal, a science gateway that integrates a suite of different bioinformatics applications using HPC and data management resources provided by the Brazilian National HPC System (SINAPAD). BioinfoPortal follows the Software as a Service (SaaS) model and the web server is freely available for academic use. The goal of this paper is to describe the science gateway and its usage, addressing challenges of designing a multiuser computational platform for parallel/distributed executions of large-scale bioinformatics applications using the Brazilian HPC resources. We also present a study of performance and scalability of some bioinformatics applications executed in the HPC environments and perform machine learning analyses for predicting features for the HPC allocation/usage that could better perform the bioinformatics applications via BioinfoPortal

Time series prediction has been attracting interest of researchers due to its increasing importance in decision-making activities in many fields of knowledge. The demand for better accuracy in time series prediction furthered the arising of many machine learning time series prediction methods (MLM). Choosing a suitable method for a particular dataset is a challenge and demands established benchmark methods (BM) for performance assessment. Suppose a particular BM is selected, and an experimental comparison is made with a particular MLM. If the latter does not provide better prediction results for the same dataset, this indicates that some improvements are needed for the MLM. Regarding this matter, adopting a well-established, easy to interpret, and tuned BM is desirable. This paper presents a framework for systematic benchmarking some MLM against well-known Linear Methods (LM), namely Polynomial Regression and models in the ARIMA family, used as BM for univariate time series prediction. We implemented such a framework within the R-Package named TSPred. This implementation was evaluated using a wide number of datasets from past prediction competitions. The results show that fittest LM provided by TSPred are adequate BM for univariate time series predictions.

A ascite quilosa é uma complicação rara após procedimentos cirúrgicos e trauma abdominal, apresentando elevada morbidade e difícil manejo. Nos casos refratários ao tratamento clínico habitual, o tratamento cirúrgico se impõe, apesar da baixa taxa de sucesso. Dois casos são apresentados: o primeiro paciente foi vítima de trauma abdominal contuso e o segundo foi submetido a hernioplastia hiatal a Nissen videolaparoscópica, ambos evoluindo com ascite quilosa que necessitou de tratamento cirúrgico através da ligadura do ducto torácico, junto aos pilares diafragmáticos. Os pacientes evoluíram com melhora clínica e ausência de ascite após 24 meses de seguimento.

International audience; Analyzing big data requires the support of dataflows with many activities to extract and explore relevant information from the data. Recent approaches such as Pig Latin propose a high-level language to model such dataflows. However, the dataflow execution is typically delegated to a MapReduce implementation such as Hadoop, which does not follow an algebraic approach, thus it cannot take advantage of the optimization opportunities of PigLatin algebra. In this paper, we propose an approach for big data analysis based on algebraic workflows, which yields optimization and parallel execution of activities and supports user steering using provenance queries. We illustrate how a big data processing dataflow can be modeled using the algebra. Through an experimental evaluation using real datasets and the execution of the dataflow with Chiron, an engine that supports our algebra, we show that our approach yields performance gains of up to 19.6% using algebraic optimizations in the dataflow and up to 39.1% of time saved on a user steering scenario.

In Cloud Computing, the utilization of virtual machines (VMs) as a computational resource unit is a fundamental concept. In a Cloud environment, VMs are deployed on physical resources with a set computational characteristics, such as processing power. VMs have computational properties as well, which are drawn from their respective hosts, and may be migrated from one host to another. In this context, the perceived performance of a parallel application is determined by how the respective VMs are allocated to the physical resources. For instance, two VMs deployed in the same host may have their performance altered in complex ways, due to resource contention. Also, the efficiency of a application that spans many VMs is dependent on the distribution of the correspond ant processes, due to available communication bandwidth between VMs, but also due to the possibility of leveraging local communication between processes or threads in the shared memory of a single host. As an attempt to describe the complex conditions that influence the performance of applications running on the same Cloud environment, we propose a more detailed representation of VM allocation, with features that can be used in the future for building more accurate performance models.

New instruments and techniques used in capturing scientific data are exponentially increasing the volume of data consumed by insilico research, which has been usually referred to as data deluge. Once captured, scientific data goes through a cleaning workflow before getting ready for analysis that will eventually confirm the scientist's hypothesis. The whole process is, nevertheless, complex and takes the focus of the scientist's attention away from his/her research and towards solving the complexity associated with managing computing products. Moreover, as the research evolves, references to previous results and workflows are needed as source of provenance data. Based on these observations, we claim that in-silico experiments must be supported by a hypotheses data model that describes the elements involved in a scientific exploration and supports hypotheses assessment. Adopting a data perspective to represent hypotheses allow high-level references to experiments and provides support for hypotheses evolution. The data model drives the proposal of a data management system that would support scientists in describing, running simulations and interpreting their results.

Several applications require reasoning over autonomously developed ontologies. Initially conceived to explicit the semantics of a certain domain, these ontologies become a powerful tool for supporting business interactions, once heterogeneities have been solved and inconsistencies eliminated. Unfortunately, a stable coherent logical state is hard to maintain in such an environment, due to normal evolution carried out independently over individual ontologies. As a result, reasoning over autonomously developed ontologies has to face with both heterogeneity and inconsistency, in order to assure correct answering. In this paper we study the problem arising in these settings. We propose an incremental reasoning approach based on a virtual reasoning space that is filled with relevant ontology entities as query answering progress. We show how to identify the set of relevant entities with respect to a user query using a set theory approach and illustrate the solution with a use case exploring the web service discovery scenario.

This paper describes an ontology-based knowledge system for creating and sharing user's personal objects, with support for querying semantic associations. The user specifies and populates a local ontology, which helps him to organize his personal objects. For users who wish to work cooperatively, or divulgate their personal objects in the so-called community of practices, the system provides a mapping between the local and global ontologies, both defined using the Resource Description Framework (RDF) model. This global ontology is available on a Web server and provides global semantic query capability. The mapping task, as well as the objects organization process, is supported by a thesaurus. It ensures that each user's view is preserved, as he is able to visualize and query objects using his original vocabulary, independently of the way they have been represented in the global ontology.

This paper describes QEF a query evaluation framework designed to support complex applications on the grid. QEF has been extended to support querying within a number of different applications, including supporting scientific visualization and implementing a web service semantic search engine. Application requests take a form of a workflow in which tasks are represented as algebraic operators and specific data types are enveloped into a common tuple structure. The implemented system is automatically deployed into schedule grid nodes and autonomously manages query evaluation according to grid environment conditions. The generality of our approach has been tested with a number of applications leading to a full grid web service implementation available at http://codims. epfl. ch.