Abstract: During the last decade, the fast evolution in communication networks has facilitated the development of complex applications that manage vast amounts of data, like Big Data applications. Unfortunately, the high complexity of these applications hampers the testing process. Moreover, generating adequate test suites to properly check these applications is a challenging task due to the elevated number of potential test cases. Mutation testing is a valuable technique to measure the quality of the selected test suite that can be used to overcome this difficulty. However, one of the main drawbacks of mutation testing lies on the high computational cost associated to this process. In this paper we propose a dynamic distributed algorithm focused on HPC systems, called EMINENT, which has been designed to face the performance problems in mutation testing techniques. EMINENT alleviates the computational cost associated with this technique since it exploits parallelism in cluster systems to reduce the final execution time. In addition, several experiments have been carried out on three applications in order to analyse the scalability and performance of EMINENT. The results show that EMINENT provides an increase in the speed-up in most scenarios.

Abstract: Finite-difference frequency-domain (FDFD) analysis of wave optics and photonics requires linear system solver for discretized vector Helmholtz equation. The linear system can be ill-conditioned when computation domain is large or perfectly-matched layers (PMLs) are used. Direct factorization of the linear systems for 3D photonic simulation may require tremendous amount of computation resources. We propose compressed hierarchical Schur method (CHiS) for computation time and memory usage savings. The results show that the CHiS method takes 45% less factorization time and 35% less memory usage compared with the uncompressed hierarchical Schur method in selected test. The computing procedure also involves many dense linear algebra operations, which can be efficiently executed in modern high-performance hardwares such as graphics processing units (GPUs) and multicore/manycore processors. We investigate the GPU acceleration strategy and hardware tuning by rescheduling the factorization. The proposed CHiS is also tested on a dual-GPU server for performance analysis. These new techniques can efficiently utilize modern high-performance environment and greatly accelerate development of future development of photonic devices and circuits.

Abstract: Combining powerful parallel frameworks and on-demand commodity hardware, cloud computing has made both analytics and decision support systems canonical to enterprises of all sizes. Associated with unprecedented volumes of data stacked by such companies, filtering and retrieving them are pressing challenges. This data is often organized in star schemas, in which Star Joins are ubiquitous and expensive operations. In particular, excessive disk spill and network communication are tight bottlenecks for all current MapReduce or Spark solutions. Here, we propose two efficient solutions that drop the computation time by at least 60%: the Spark Bloom-Filtered Cascade Join (SBFCJ) and the Spark Broadcast Join (SBJ). Conversely, a direct Spark implementation of a sequence of joins renders poor performance, showcasing the importance of further filtering for minimal disk spill and network communication. Finally, while SBJ is twice faster when memory per executor is large enough, SBFCJ is remarkably resilient to low memory scenarios. Both algorithms pose very competitive solutions to Star Joins in the cloud.

Abstract: High Performance Computing has traditionally been the natural habitat of highly specialized parallel programming experts running large batch jobs. With every field of Science becoming richer and richer in the amount of data available, many more scientists are transitioning to Supercomputers or cloud computing resources. In this paper I would like to review how the Jupyter project, a suite of scientific computing tools, can help to democratize access to Supercomputers by lowering the entry barrier for new scientific communities and provide a gradual path to harnessing more distributed computing capabilities. I will start from the interactive usage of the Jupyter Notebook, a widespread browser-based data exploration environment, on a HPC cluster, then explain how notebooks can be used as scripts directly or in a workflow environment and finally how batch data processing like traditional MPI, Spark and XSEDE Gateways can benefit from inter-operating with a Jupyter Notebook environment.

Abstract: LDA is a widely used machine learning technique for big data analysis. The application includes an inference algorithm that iteratively updates a model until it converges. A major challenge is the scaling issue in parallelization owing to the fact that the model size is huge and parallel workers need to communicate the model continually. We identify three important features of the model in parallel LDA computation: 1. The volume of model parameters required for local computation is high; 2. The time complexity of local computation is proportional to the required model size; 3. The model size shrinks as it converges. By investigating collective and asynchronous methods for model communication in different tools, we discover that optimized collective communication can improve the model update speed, thus allowing the model to converge faster. The performance improvement derives not only from accelerated communication but also from reduced iteration computation time as the model size shrinks during the model convergence. To foster faster model convergence, we design new collective communication abstractions and implement two Harp-LDA applications, "lgs" and "rtt". We compare our new approach with Yahoo! LDA and Petuum LDA, two leading implementations favoring asynchronous communication methods in the field, on a 100-node, 4000-thread Intel Haswell cluster. The experiments show that "lgs" can reach higher model likelihood with shorter or similar execution time compared with Yahoo! LDA, while "rtt" can run up to 3.9 times faster compared with Petuum LDA when achieving similar model likelihood.

Abstract: Despite being widely visible on the web, Internet-promoted commercial sex work has so far attracted limited attention from the side of researchers. Current studies outline the issues that new forms of sex work are associated with, however, very little is known to date about their spatial manifestation. In this research we follow the environmental perspective in spatial analysis of crime and deviance with the assumption that the location of venues for provision of commercial sex work can be modeled via the algorithms trained on the distribution of possible correlates in the proximity to the existing venues. Visualization of the acquired results is presented herein along with the errors and score metrics for evaluation of the applicability of specific methods of machine learning. The paper is concluded with the estimation of potential extensions and peculiarities of data used in the research.

Abstract: Understanding retweeting mechanism and predicting retweeting behavior is an important and valuable task in user behavior analysis. In this paper, aiming at providing a general method for improving retweeting behavior prediction performance, we propose a probabilistic matrix factorization model (RTPMF) incorporating user social network information and message semantic relationship. The contributions of this paper are three-fold: (1) We convert predicting user retweeting behavior problem to solve a probabilistic matrix factorization problem; (2) Following the intuition that user social network relationship will affect the retweeting behavior, we extensively study how to model social information to improve the prediction performance; and (3) We also incorporate message semantic embedding to constrain the objective function by making a full use of additional the messages' content-based and structure-based features. The empirical results and analysis demonstrate that our method significantly outperform the state-of-the-art approaches.

Abstract: Sentiment analysis is concerned with classifying a subjective text into positive or negative according to the opinion expressed in it. The performance of traditional sentiment classification algorithms rely heavily on manually labeled training data. However, not every domain has the labeled data because the labeling work is time-consuming and expensive. In this paper, we propose a latent sentiment factorization (LSF) algorithm based on probabilistic matrix factorization technique for cross-domain sentiment classification. LSF works in the setting where there are only labeled data in the source domain and unlabeled data in the target domain. It bridges the gap between domains by exploiting the sentiment correlations between domain-shared and domain-specific words in a two-dimensional sentiment space. Experimental results demonstrate the superiority of our method over the state-of-the-art approaches.

Abstract: Identifying users across different sites is to find the accounts that belong to the same individual. The problem is fundamental and important, and its results can benefit many applications such as social recommendation. Observing that 1) usernames are essential elements for all sites; 2) most users have limited number of usernames on the Internet; 3) usernames carries information that reflect an individual’s characteristics and habits etc., this paper tries to identify users based on username similarity. Specifically, we introduce the self-information vector model to integrate our proposed content and pattern features extracted from usernames into vectors. In this paper, we define two usernames’ similarity as the cosine similarity between their self-information vectors. We further propose an abbreviation detection method to discover the initialism phenomenon in usernames, which can improve our user identification results. Experimental results on real-world username sets show that we can achieve 86.19% precision rate, 68.53% recall rate and 76.21% F1-measure in average, which is better than the state-of-the-art work.

Abstract: A wealth of valuable data is locked within the millions of research articles published each year. Reading and extracting pertinent information from those articles has become an unmanageable task for scientists. This problem hinders scientific progress by making it hard to build on results buried in literature. Moreover, these data are loosely structured, encoded in manuscripts of various formats, embedded in different content types, and are, in general, not machine accessible. We present a hybrid human-computer solution for semi-automatically extracting scientific facts from literature. This solution combines an automated discovery, download, and extraction phase with a semi-expert crowd assembled from students to extract specific scientific facts. To evaluate our approach we apply it to a particularly challenging molecular engineering scenario, extraction of a polymer property: the Flory-Huggins interaction parameter. We demonstrate useful contributions to a comprehensive database of polymer properties.

Abstract: Following recently published socio-cognitively inspired ACO concept for global optimization, we try to verify the proposed idea by adapting the PSO in a similar way. The swarm is divided into species and the particles get inspired not only by the global and local optima, but share the knowledge about the optima with neighbourhood agents belonging to other species. After presenting the concept and motivation, the experimental results gathered for common benchmark functions tackled in 100 dimensions are shown and the efficacy of the proposed algorithm is discussed.

Abstract: In swarm robotics, it is necessary to develop methods and strategies that guide the collective execution of tasks by the robots. The design of such tasks can be done considering it as a collection of simpler behaviors, called subtasks. In this paper, the Wave Swarm is presented as a general strategy to manage the sequence of subtasks that compose the collective navigation , which is an important task in swarm robotics. The proposed strategy is based mainly on the execution of wave algorithms. The swarm is viewed as a distributed system, wherein the communication is achieved by message passing among robot’s neighborhood. Message propagation delimits the start and end of each subtask. Simulations are performed to demonstrate that controlled navigation of robot swarms/clusters is achieved with three subtasks, which are recruitment, alignment and movement.

Abstract: The growing share of LED light sources in outdoor lighting enables developing street lighting solutions characterized by high energy efficiency. It is accomplished by replacing high intensity discharging lamps with LEDs and implementing various control strategies. It was also shown that a well tailored lighting design may significantly decrease the power usage. To apply this method in large projects, however, the computationally efficient approach is necessary. In this article we propose the method of energy efficiency optimization relying on a multi-agent system framework which enables scalable computations capable of handling large-scale projects. The case of a real-life optimization is also presented in the paper.

Abstract: We study the features of a new mixed integration scheme dedicated for solving the nonstationary variational problems. The scheme is composed of the FEM approximation with respect to the space variable coupled with a 3-layered time integration scheme with a linearized right-hand side operator. It was applied in solving the Cahn-Hilliard parabolic equation with a nonlinear, fourth-order elliptic part. The second order of the approximation along the time variable was proven. Moreover, the good scalability of the software based on this scheme was confirmed during simulations. We verify the proposed time integration scheme by monitoring the Ginzberg-Landau free energy. The numerical simulations are performed using parallel multifrontal direct solver executed over STAMPEDE Linux cluster. Its scalability was compared to the results of the three direct solvers, including MUMPS, SuperLU and PaSTiX.

Abstract: Kernels in the ADER-DG method, when solving the elastic wave equations, boil down to sparse and dense matrix multiplications of small sizes. At the example of the Earthquake simulations code SeisSol, we will investigate how this routines can be implemented and speeded-up by the code generation tool LIBXSMM. A long these lines we will analyze different tradeoffs of switching from sparse to dense matrix multiplication kernels and report performance with respect to time-to-solution and energy consumption. Bio: Alexander Heinecke studied Computer Science and Finance and Information Management at Technische Universität München, Germany. In 2010 and 2012, he completed internships at Intel in Munich, Germany and at Intel Labs Santa Clara, CA, USA. In 2013 he completed his Ph.D. studies at TUM and joined Intel’s Parallel Computing in Santa Clara in 2014. His core research topic is the use of multi- and many-core architectures in advanced scientific computing applications. In 2014, he and his co-authors were selected as Gordon Bell finalists for running multi-physics earthquake simulations at multi-petaflop performance on more than 1.5 million of cores.

Abstract: We developed an unstructured finite element mesh generation method capable of modeling multiple-material complex geometry problems for large-scale seismic analysis of soil-structure systems. We used an octree structure to decompose the target domain into small subdomains and use the multiple material marching cubes method for robust and parallel tetrahedralization of each subdomain. By using the developed method on a 32 core shared memory machine, we could generate a 594,168,792 tetrahedral element soil-structure model of a power plant in 13 h 01 min. The validity of the generated model was confirmed by conducting a seismic response analysis on 2,304 compute nodes of the K computer at RIKEN. Although the model contains a small approximation in geometry (half of the minimum octree size) at present, we can expect fast and high quality meshing of large-scale models by making geometry correction in the future, which is expected to help improve the seismic safety of important structures and complex urban systems.

Abstract: The efficiency and robustness of preconditioned parallel iterative solvers, based on domain decomposition for ill-conditioned problems with heterogeneous material properties, are evaluated in the present work. The preconditioning method is based on the BILUT(p,d,t) method proposed by the author in a previous study, and two types of domain decomposition procedures, LBJ (Localized Block Jacobi) and HID (Hierarchical Interface Decomposition), are considered. The proposed methods are implemented using the Hetero3D code, which is a parallel finite-element benchmark program for solid mechanics problems, and the code provides excellent scalability and robustness on up to 240 nodes (3,840 cores) of the Fujitsu PRIMEHPC FX10 (Oakleaf-FX) at the Information Technology Center, the University of Tokyo. Generally, HID provides better efficiency and robustness than LBJ for a wide range of values of parameters.

Abstract: Pricing and hedging exotic options using local stochastic volatility models drew a serious attention within the last decade, and nowadays became almost a standard approach to this problem. In this paper we show how this framework could be extended by adding to the model stochastic interest rates and correlated jumps in all three components. We also propose a new fully implicit modification of the popular Hundsdorfer and Verwer and Modified Craig-Sneyd finite-difference schemes which provides second order approximation in space and time, is unconditionally stable and preserves positivity of the solution, while still has a linear complexity in the number of grid nodes.

Abstract: We will present a method to numerically price options by solving the Fokker-Planck equation for the conditional probability density p(s,t|s_0,t_0). This enables the pricing of several contracts with pay-offs ϕ(s,K,T) (with strike-price K and time of maturity T) by integrating p(s,T|s_0,t_0) multiplied by ϕ(s,K,T) and discount to today's price. From a numerical perspective the initial condition for the Fokker-Planck equation is particularly challenging since it is a Dirac delta function. In [1] a closed-form expansion for the conditional probability density was introduced that is valid for small time-steps. We use this for the computation of p(s,t_0+∆t|s_0,t_0) the first time-step. For the remaining time-steps we discretize the Fokker-Planck equation using BDF-2 in time and Radial Basis Function (RBF) approximation in space with Gaussian RBFs. Finally, the computation of the option prices from the obtained p(s,T|s_0,t_0) can be done analytically for many pay-off functions ϕ(s,K,T), due to the Gaussian RBFs. We will demonstrate the good qualities of our proposed method for European call options and barrier options. [1] Y. Aït-Sahalia, Maximum-likelihood estimation of discretely-sampled diffusions: A closed-form approximation approach, Econometrica, 70: 223–262, 2002.

Abstract: Tail dependence refers to clustering of extreme events. In the context of financial risk management, the clustering of high-severity risks has a devastating effect on the well-being of firms and is thus of pivotal importance in risk analysis. When it comes to quantifying the extent of tail dependence, it is generally agreed that measures of tail dependence must be independent of the marginal distributions of the risks but rather solely copula-dependent. Indeed, all classical measures of tail dependence are such, but they investigate the amount of tail dependence along the main diagonal of copulas, which has often little in common with the concentration of extremes in the copulas' domain of definition. In this paper we urge that the classical measures of tail dependence may underestimate the level of tail dependence in copulas. For the Gaussian copula, however, we prove that the classical measures are maximal. As, in spite of the numerous criticisms, the Gaussian copula remains ubiquitous in a great variety of practical applications, our ndings must be a welcome news for risk professionals.

Abstract: A radial basis function generated finite difference (RBF-FD) method has been considered for solving multidimensional PDEs arising in pricing of financial contracts, mainly basket options. Being mesh-free while yielding a sparse differentiation matrix, this method aims to exploit the best properties from, both, finite difference (FD) methods and radial basis function (RBF) methods. Moreover, the RBF-FD method is expected to be advantageous for high-dimensional problems compared to: Monte Carlo (MC) methods which converge slowly, global RBF methods since they produce dense matrices, and FD methods because they require regular grids. The method was succesfully tested in solving the standard Black-Scholes-Merton equation for pricing European and American options with discrete or continuous dividends in 1D. Then, it is developed further in order to price European call basket and spread options in 2D on adapted domains, and some groundwork has been done in solving 3D problems as well. The method features a non-uniform node placement in space, as well as a variable spatial stencil size, in order to improve the accuracy in the regions with known low regularity. Performance of the method and the error profiles have been studied with respect to discretization in space, size and form of stencils, and RBF shape parameter. The results highlight RBF-FD as a competitive, sparse method, capable of achieving high accuracy with a small number of nodes in space.

Abstract: Credit risk models largely bifurcate into two classes – the structural models and the reduced form models. Attempts have been made to reconcile the two approaches via restricting information by adjusting filtrations, but they are technically complicated. Here we propose a reconciliation inspired by actuarial science’s approach to survival analysis. Extending the work of Chen, we model the hazard rate curve itself as a stochastic process. This puts default models in a form resembling the HJM model for interest rates, yielding a unifying framework for default modeling. All credit models can be put in this form, and default dependent derivatives can be directly priced in this framework. Predictability of default has a simple interpretation in this framework. The framework enables us to disentangle predictability and the distribution of the default time from calibration decisions such as whether to use market prices or balance sheet information. It also allows us a simple way to define new default models.

Abstract: In this paper, we develop a two-scale reduced model for simulating the Darcy flow in two-dimensional porous media with conductive fractures. We apply the approach motivated by the embedded fracture model (EFM) to simulate the flow on the coarse scale, and the effect of fractures on each coarse scale grid cell intersecting with fractures is represented by the discrete fracture model (DFM) on the fine scale. In the DFM used on the fine scale, the matrix-fracture system are resolved on unstructured grid which represents the fractures accurately, while in the EFM used on the coarse scale, the flux interaction between fractures and matrix are dealt with as a source term, and the matrix-fracture system can be resolved on structured grid. The Raviart-Thomas mixed finite element methods are used for the solution of the coupled flows in the matrix and the fractures on both fine and coarse scales. Numerical results are presented to demonstrate the efficiency of the proposed model for simulation of flow in fractured porous media.

Abstract: Reduced-order model by proper orthogonal decomposition of Navier-Stokes equation can be established in different manners. After careful screening under different sampling intervals and numbers of basis vectors, it has been found that the model can achieve high precision only when it is constructed on collocated grid with the samples still on the staggered grid. The model straight-forward established on the staggered grid may lose accuracy apparently. To precisely capture the dynamic behavior of flow field, the sampling interval should be small enough while the number of basis vectors should be moderate. These conclusions can be a valuable principle for future modeling of the dynamics of fluid flow.

Abstract: In this paper, we introduce a mathematical model to describe the nanoparticles transport carried by a two-phase flow in a porous medium including gravity, capillary forces and Brownian diffusion. Nonlinear iterative IMPES scheme is used to solve the flow equation, and saturation and pressure are calculated at the current iteration step and then the transport equation is solved implicitly. Therefore, once the nanoparticles concentration is computed, the two equations of volume of the nanoparticles available on the pore surfaces and the volume of the nanoparticles entrapped in pore throats are solved implicitly. The porosity and the permeability variations are updated at each time step after each iteration loop. Numerical example for regular heterogenous permeability is considered. We monitor the changing of the fluid and solid properties due to adding the nanoparticles. Variation of water saturation, water pressure, nanoparticles concentration and porosity are presented graphically.

Abstract: In this work, an efficient coupling between Monte Carlo (MC) molecular simulation and Darcy-scale flow in porous media is presented. The cell centered finite difference method with non-uniform rectangular mesh were used to discretize the simulation domain and solve the governing equations. To speed up the MC simulations, we implemented a recently developed scheme that quickly generates MC Markov chains out of pre-computed ones, based on the reweighting and reconstruction algorithm. This method astonishingly reduces the required computational times by MC simulations from hours to seconds. To demonstrate the strength of the proposed coupling in terms of computational time efficiency and numerical accuracy in fluid properties, various numerical experiments covering different compressible single-phase flow scenarios were conducted. The novelty in the introduced scheme is in allowing an efficient coupling of the molecular scale and the Darcy's one in reservoir simulators. This leads to an accurate description of thermodynamic behavior of the simulated reservoir fluids; consequently enhancing the confidence in the flow predictions in porous media.

Abstract: This research addresses a sequential convex splitting method for numerical simulation of multicomponent two-phase fluids mixture in a single-pore at constant temperature, which is modeled by the gradient theory with Peng-Robinson equation of state. The gradient theory of thermodynamics and variational calculus are utilized to obtain a system of chemical equilibrium equations which are transformed into a transient system as a numerical strategy on which the numerical scheme is based. The proposed numerical algorithm avoids computing Hessian matrix arising from the second-order derivative of homogeneous contribution of free energy; it is also quite robust. This scheme is proved to be unconditionally component-wise energy stable. The Raviart-Thomas mixed finite element method is applied to spatial discretization.

Abstract: Prof. Angela Shiflet in computer science and mathematics and Prof. George Shiflet in biology are Fulbright Specialists. In January, 2015, they participated in a three-week collaborative project at University “Magna Græcia” of Catanzaro in Italy, in the Department of Medical and Surgical Sciences, hosted by Prof. Mario Cannataro. While there, the three along with Prof. Pietro Hiram Guzzi started a project to develop educational module(s) on one or more high-performance-computing bioinformatics algorithms. Drs. Cannataro and Guzzi have written a book, Data Management of Protein Interaction Networks (Wiley, 2011), and regularly teach bioinformatics and HPC. Upon returning to the United States, the Drs. Shiflet applied to have undergraduate Daniel Couch be a Blue Waters Intern for one year working on the project. The NSF-funded Blue Waters Project, which provides a stipend for the intern, supports “experiences involving the application of high-performance computing to problems in the sciences, engineering, or mathematics” (http://computationalscience.org/bwsip/). Besides having had an HPC course, the student participated in a two-week workshop at the National Center for Supercomputing Applications (NCSA) facilities on the University of Illinois Urbana-Champaign campus. In the project, he has written sequential and HPC programs and performed timings to accompany an educational module on “Aligning SequencesSequentially and Concurrently,” available at http://www.wofford.edu/ecs/, and is working with the professors on developing other modules and programs. After covering the necessary biological background, the named module develops the sequential Needleman-Wunsch Algorithm (NWA) to determine the similarity and the alignment(s) that yield a highest similarity score. Employing timings developed by the intern, the module illustrates that the algorithm’s runtime is proportional to the square of the number of nucleotides. Having motivated the need for HPC, the module discusses HPC pipeline versions of NWA along with timings. To aid students, the module contains fifteen Quick Review Questions, many with multiple parts; nine exercises; and five projects. Completed sequential and parallel C with MPI programs are available upon request by instructors. The materials are current being used by students and faculty members in a bioinformatics course at University “Magna Græcia” of Catanzaro.

Abstract: This paper presents an experience of an introductory course on Parallel Programming. The course is dedicated to parallel programming tools and environments, and in particular to the analysis, development and optimization of parallel algorithms. It has a practical orientation and is guided with the use of problems from the Spanish Parallel Programming Contest. The different units are presented in the traditional lecture format, and a practical session accompanies each unit, with problems to work with in the tools or algorithmic paradigms presented in the previous lecture. The students work in the practical sessions on problems and using the system of the contest, which facilitates online and real time validation of their implementations. The practical approach of the course and the continuous evaluation used led to an important increase in the marks.

Abstract: In this talk we discuss designing, implementing, and researching an Introduction to Computational Modeling course for university undergraduates. The course is part of the brand-new Computational Mathematics, Science, and Engineering (CMSE) department at Michigan State University (MSU). It was specifically designed to be interdisciplinary; the course is open to any major at the university. The course was also built to address the growing need for a workforce that can analyze, model, and interpret real-world data. Our talk will cover three strands of developing the course: 1. Curriculum Design: how we worked backwards from the professional disciplinary practices of modeling to arrive at well-defined learning outcomes, assessments, and course content. 2. Instructional Environment: The key decisions we made and technologies we chose to bring the experience of modeling to the classroom. 3. Educational Research: How we’re using methods from the learning sciences—including clinical interviews and ethnographic classroom observation—to understand students’ experiences in the course and continually integrate findings into the course design

Abstract: Each scientific project or publication can be attributed to several fields of study with different degrees. Call interdisciplinary distribution the set of the degrees. If we have fixed number of fields of study, the set can be written in vector form. Each component of the vector corresponds to one of the fields of study. Call the vector the interdisciplinary vector. If we consider a scientist to be a set of his or her publications we can get the vector for a scientist as a weighted sum of a vector of his or her publications. This paper is devoted to an approach to the evaluation of the interdisciplinary distribution of professional or research objects (RO), and the transdisciplinary effects of their changes. RO and professionals can be evaluated on the basis of keywords in relevant scientific papers, reports, surveys, proposals, CVs, and so on. The transdisciplinary effect is apparent when the interdisciplinarity distribution has been changed. We propose formulas to evaluate this transdisciplinary effect. This approach was implemented using participants in group projects at the fourth Young Scientists Conference (YSC) on High-Performance Computing and Computer Simulation. The accuracy of the interdisciplinary vector of several participants was examined by the survey about their involvement in the team projects. This approach can be used to evaluate the compliance of a scientific team with the transdisciplinary research project (problem), as well as to assess the students' skills in transdisciplinary environments.

Abstract: We are reporting on a strong capacity boost in storing digital data in synthetic DNA. In principle, synthetic DNA is an ideal media to archive digital data for very long times because the achievable data density and longevity outperforms today’s digital data storage media by far. On the other hand, neither the synthesis, nor the amplification and the sequencing of DNA strands can be performed error-free today and in the foreseeable future. In order to make synthetic DNA available as digital data storage media, forward-error-correction schemes have to be applied. In order to realize DNA data storage, we have developed an efficient and robust forward-error-correcting scheme adapted to the DNA channel. We based the design of the needed DNA channel model on data from a proof-of-concept conducted 2012 by a team from the Harvard Medical School*. Our forward error correction scheme is able to cope with all error types of today DNA synthesis, amplification and sequencing processes, e.g. insertion, deletion, and swap errors. In a successful experiment, we were able to store and retrieve error-free 22MByte of digital data in synthetic DNA recently. The found residual error probability is already in the same order as it is in hard disk drives and can be easily improved. This proves the feasibility to use synthetic DNA as a long-term digital data storage media. In an already planned next development step we will increase the amount of stored data into the GByte range. The presented forward error correction scheme is already designed for such and even much higher volumes of data. *) Church, G. M.; Gao, Y.; Kosuri, S. (2012). "Next-Generation Digital Information Storage in DNA". Science 337 (6102): 1628

Abstract: In this paper, we describe some available high-confident call sets that have been developed to test the accuracy of called single nucleotide polymorphisms (SNPs) from next-generation sequencing. We use these calls to test and parameterize the GATK best practice pipeline on the high-performance computing cluster at the University of Kentucky. Automated script to run the pipeline can be found at https://github.com/sallyrose0425/GATKBP. This study demonstrates the usefulness of high-confident call sets in validating and optimizing bioinformatics pipelines, estimates computational needs for genomic analysis, and provides scripts for an automated GATK best practices pipeline.

Abstract: There are many legitimate reasons why standards for formatting of biomedical research data are lengthy and complex (Souza, Kush, & Evans, 2007). However, the common scenario of a biostatistician simply needing to import a given dataset into their statistical software is at best under-served by these standards. Statisticians are forced to act as amateur database administrators to pivot and join their data into a usable form before they can even begin the work that they specialize in doing. Or worse, they find their choice of statistical tools dictated not by their own experience and skills, but by remote standards bodies or inertial administrative choices. This may limit academic freedom. If the formats in question require the use of one proprietary software package, it also raises concerns about vendor lock-in (DeLano, 2005) and stewardship of public resources. The logistics and transparency of data sharing can be made more tractable by an appreciation of the differences between structural, semantic, and syntactic levels of data interoperability. The semantic level is legitimately a complex problem. Here we make the case that, for the limited purpose of statistical analysis, a simplifying assumption can be made about structural level: the needs of a large number of statistical models can often be met with a modified variant of the first normal form or 1NF (Codd, 1979). Once data is merged into one such table, the syntactic level becomes a solved problem, with many text based formats available and robustly supported by virtually all statistical software without the need for any custom or third-party client-side add-ons. We implemented our denormalization approach in DataFinisher, an open source server-side add-on for i2b2 (Murphy et al., 2009), which we use at our site to enable self-service pulls of de-identified data by researchers.

Abstract: Online users may use many different channels, devices and venues for any online user experience. To make all services such as web design, ads, web content, shopping, personalized for every user; we need to be able to recognize them regardless of device, channels and venues they are using. This, in turn, requires building up a comprehensive view of the user which includes all of their behavioral characteristics - that are spread all over these different venues. This would not be possible without having all behavioral related data of the user which requires the capacity of connecting the user all over the devices, and channels, so to have all of their behavior under a single view. This work is a major attempt in doing this using only behavioral data of users while protecting the user’s privacy.

Abstract: Current recommender systems are very inefficient. There are many metrics that are used to measure the effectiveness of recommender systems. These metrics often include “conversion rate” and “click through rate”. Recently, these rates are in low single digit (less than 10%). In other words, more than 90% of times, the model that the targeting system is based on, produces noise. The belief in this work is that the main problem leading to getting such unsatisfactory outcomes is the modeling problem. Much of the modeling problem could be represented and exemplified in treating users and items as member of clusters (segments). In this work, we consider full personalization of recommendation systems. We aim at personalization of users and items simultaneously.

Abstract: We present a method of generating item recommendations by learning item feature vector embeddings. Our work is analogous to approaches like Word2Vec or Glove used to generate a good vector representation of words in a natural language corpus. We treat the items that a user interacted with as analogous to words and the string of items interacted with in a session as sentences. Our embedding generates semantically related clusters and the item vectors generated can be used to compute item similarity which can be used to drive product recommendations. Our method also allows us to use the feature vectors in other machine learning systems. We validate our method on the MovieLens dataset.

Abstract: The default decision tree builder in Mahout 0.9 has severe implementation problems that build small, weak decision trees which limit its usefulness in production situations when the features are strictly numerical. In this talk I will describe a simple, more powerful decision tree builder that systematically produces regression models with much better AUCs without sacrificing performance. The new builder also creates models that are of relatively compact size (about 30-50 Kb in the tested data sets), as compared to the large (500 Kb – 2 Mb) models generated from a fixed version of the original decision tree builder. I will describe the problem with the Mahout decision tree builder and the simple replacement and how they work, and will compare the model size, build times, and AUC performance on several historic data sets from Adobe Target from customers in different industries to show that the improvement is very general.

Abstract: Computationally ecient sensitivity analysis of a large-scale air pol- lution model is an important issue we focus on in this paper. Sensitivity studies play an important role for reliability analysis of the results of com- plex nonlinear models as those used in the air pollution modelling. There is a number of uncertainties in the input data sets, as well as in some internal coecients, which determine the speed of the main chemical re- actions in the chemical part of the model. These uncertainties are subject to our quantitative sensitivity study. Monte Carlo and quasi-Monte Carlo algorithms are used in this study. A large number of numerical experiments with some special modica- tions of the model must be carried out in order to collect the necessary input data for the particular sensitivity study. For this purpose we cre- ated an ecient high performance implementation SA-DEM, based on the MPI version of the package UNI-DEM. A large number of numerical ex- periments were carried out with SA-DEM on the IBM MareNostrum III at BSC - Barcelona, helped us to identify a severe performance problem with an earlier version of the code and to resolve it successfuly. The im- proved implementation appears to be quite ecient for that challenging computational problem, as our experiments show. Some numerical results with performance and scalability analysis of these results are presented in the paper.

Abstract: -