Keynote Speakers
Inderjit S Dhillon
Gottesman Family Centennial Professor
University of Texas at Austin
Department of Computer Science
Austin, TX, USA
He is an ACM Fellow, IEEE Fellow, SIAM Fellow, AAAS Fellow and has won the SIAM Linear Algebra Prize. His research interests are in machine learning, large-scale data analysis and bioinformatics. His emphasis is on developing novel algorithms that respect the underlying problem structure and are scalable to massive data sets. Some of his current research topics: high-dimensional data analysis, divide-and-conquer methods for big data analytics, social network analysis, and predicting gene-disease associations.

Title of Talk

Stabilizing Gradients for Deep Neural Networks

Vanishing and exploding gradients are two of the main obstacles in training deep neural networks, especially in capturing long range dependencies in recurrent neural networks (RNNs). In this talk, we present an efficient parametrization of the transition matrix of an RNN that allows us to stabilize the gradients that arise in its training. Specifically, we parameterize the transition matrix by its singular value decomposition (SVD), which allows us to explicitly track and control its singular values. We attain efficiency by using tools that are common in numerical linear algebra, namely Householder reflectors for representing the orthogonal matrices that arise in the SVD. By explicitly controlling the singular values, our proposed svdRNN method allows us to easily solve the exploding gradient problem and we observe that it empirically solves the vanishing gradient issue to a large extent. We note that the SVD parameterization can be used for any rectangular weight matrix, hence it can be easily extended to any deep neural network, such as a multi-layer perceptron. Theoretically, we demonstrate that our parameterization does not lose any expressive power, and show how it potentially makes the optimization process easier. Our experimental results demonstrate that the proposed framework converges faster, and has better generalization, especially when the depth is large.

David A. Bader
Professor and Chair,
School of Computational Science
and Computing Georgia Institute of Technology
Brief Speaker Bio :
David A. Bader is Professor and Chair of the School of Computational Science and Engineering, College of Computing, at Georgia Institute of Technology. He is a Fellow of the IEEE and AAAS and served on the White House's National Strategic Computing Initiative (NSCI) panel. Dr. Bader serves as a board member of the Computing Research Association, on the NSF Advisory Committee on Cyberinfrastructure, on the Council on Competitiveness High Performance Computing Advisory Committee, on the IEEE Computer Society Board of Governors, and on the Steering Committees of the IPDPS and HiPC conferences. He is the editor-in-chief of IEEE Transactions on Parallel and Distributed Systems, and is a National Science Foundation CAREER Award recipient. Dr. Bader is a leading expert in data sciences. His interests are at the intersection of high-performance computing and real-world applications, including cybersecurity, massive-scale analytics, and computational genomics, and he has co-authored over 210 articles in peer-reviewed journals and conferences. During his career, Dr. Bader has served as PI/coPI of over US$179M of competitive awards with over US$41.1M of this brought into his institution. Dr. Bader has served as a lead scientist in several DARPA programs including High Productivity Computing Systems (HPCS) with IBM PERCS, Ubiquitous High Performance Computing (UHPC) with NVIDIA ECHELON, Anomaly Detection at Multiple Scales (ADAMS), Power Efficiency Revolution For Embedded Computing Technologies (PERFECT), and Hierarchical Identify Verify Exploit (HIVE). He has also served as Director of the Sony-Toshiba-IBM Center of Competence for the Cell Broadband Engine Processor. Bader is a co-founder of the Graph500 List for benchmarking "Big Data" computing platforms. Bader is recognized as a "RockStar" of High Performance Computing by InsideHPC and as HPCwire's People to Watch in 2012 and 2014. Dr. Bader also serves as an associate editor for several high impact publications including IEEE Transactions on Computers, ACM Transactions on Parallel Computing, and ACM Journal of Experimental Algorithmics. He successfully launched his school's Strategic Partnership Program in 2015, whose partners include Accenture, Booz Allen Hamilton, Cray, IBM, Keysight Technologies, LexisNexis, Northrop Grumman, NVIDIA, and Yahoo; as well as the National Security Agency, Sandia National Laboratories, Pacific Northwest National Laboratory, and Oak Ridge National Laboratory.

Title of Talk

Massive-scale streaming analytics

Emerging real-world graph problems include: detecting community structure in large social networks; improving the resilience of the electric power grid; and detecting and preventing disease in human populations. Unlike traditional applications in computational science and engineering, solving these problems at scale often raises new challenges because of the sparsity and lack of locality in the data, the need for additional research on scalable algorithms and development of frameworks for solving these problems on high performance computers, and the need for improved models that also capture the noise and bias inherent in the torrential data streams. In this talk, the speaker will discuss the opportunities and challenges in massive data-intensive computing for applications in computational science and engineering.
Invited Speakers

Guna Seetharaman
Senior Scientist ST for Advanced Computing Concepts
Current : U.S. Naval Research Laboratory
Previous : Air Force Research Laboratory, IEEE Mohawk Valley Section, Air Force Institute of Technology
Education : University of Miami
Guna Seetharaman is the Navy Senior Scientist (ST) for Advanced Computing Concepts, and the Chief Scientist for Computation, Center for Computational Science, Navy Research Lab. He leads a team effort on: Video Analytics, High performance computing, low–latency, high-throughput, on-demand scalable geo-dispersed computer-networks. He joined NRL in June 2015. He worked as Principal Engineer at the Air Force Research Laboratory, Information Directorate, where he led research and development in Video Exploitation, Wide Area Motion Imagery, Computing Architectures and Cyber Security. He holds three US Patents, and has filed more disclosures, in related areas. His team won the best algorithm award at IEEE CVPR-2014 Video Change Detection challenge, featuring a semantic segmentation of dynamic scene to detect change in the midst of dynamic clutters. He served as a tenured professor at the Air Force Institute of Technology, and University of Louisiana at Lafayette, before joining AFRL. He and his colleagues cofounded Team CajubBot and successfully fielded two unmanned vehicles at the DARPA Grand Challenges 2004, 2005 and 2007. He also co-edited a special issue of IEEE COMPUTER dedicated to Unmanned Vehicles, and special issue of The European Journal Embedded Systems focused on intelligent autonomous vehicles. He was elected as Fellow of the IEEE, in 2014, for his contributions in high performance computer vision algorithms for airborne applications. He also served as the elected Chair of the IEEE Mohawk Valley Section, Region 1, FY 2013 and FY 2014.

Title of Talk

Computing Architectures for Machine Perception in the Post Deep-Learning World

We are in the middle of a surge in popularity, interest and investments in artificial intelligence and neural networks. Successful demonstration of ImageNet classification with deep convolutional neural networks (DCNN), has established the viability of robust image recognition system based on very large collection of labeled images, and powerful GPUS to train the DCNN, and execute the trained classifier on computing devices with modest computing capabilities. Typically the training takes place on a centralized computer, and classification takes place on computing device connected directly to the sensor, referred as computing at the edge. Elements of the analytical framework of image recognition include convolution, shift invariance, correlation and scaling by down-sampling are easier are essential to the pooling process critical for building deep neural networks. Similar abstractions for analyzing large collection of texts, and cyber forensic data, are yet to be well established. The computing capabilities required to build ubiquitous self-aware multimodal-sensor network will entail architectural innovations beyond GPGPUs. We will review the lessons learned in the realm of video analytics and automated text analytics over the past decades and also share insights towards promising new directions. Exciting new possibilities await us, enabled by the convergence of open-source software development tools, data- sharing tools, and innovatively packaged processors with high-bandwidth memory subsystems and highly scalable networked computing.

P. (Saday) Sadayappan
Department of Computer Science and Engineering
595 Dreese Lab, 2015 Neil Avenue
Ohio State University, Columbus, Ohio 43210 USA
+1-614-292-0053 (office), +1-614-292-2911 (fax)
Brief Speaker Bio :
P. (Saday) Sadayappan is a Professor of Computer Science and Engineering at The Ohio State University. His research interests include compiler optimization for heterogeneous systems, domain/pattern-specific compiler optimization, characterization of data movement complexity of algorithms, and data-structure centric performance optimization. He obtained a Bachelors degree from IIT-Madras, and an M.S. and Ph.D. from Stony Brook University. He is a Fellow of the IEEE.

Title of Talk

GPU acceleration of data/graph analytics algorithms

GPUs have considerably higher peak performance than the most powerful multicore CPUs. While GPUs are now widely used for some computations like CNNs (Convolutional Neural Networks), there are challenges in achieving high GPU performance for many data/graph analytics computations. This talk will discuss these challenges and some approaches to addressing them.

Chandrajit Bajaj
Computer Science, and Institute of Computational Engineering and Sciences,
Center for Computational Visualization,
The University of Texas - Austin
Brief Speaker Bio:
Chandrajit Bajaj is the director of the Center for Computational Visualization, in the Institute for Computational and Engineering Sciences (ICES) and a Professor of Computer Sciences at the University of Texas at Austin. Bajaj holds the Computational Applied Mathematics Chair in Visualization. He is also an affiliate faculty member of Mathematics, Electrical Engineering, Bio-Medical Engineering, Neurobiology, and a fellow of the Institute of Cell and Molecular Biology. He is currently on the editorial boards for the International Journal of Computational Geometry and Applications, and the ACM Computing Surveys, and past editorial member of the SIAM Journal on Imaging Sciences. He is a Distinguished Alumunus of the Indian Institute of Technology, Delhi, (IIT, Delhi). He is also a Fellow of The American Association for the Advancement of Science (AAAS), Fellow of the Association for Computing Machinery (ACM), Fellow of the Institute of Electrical and Electronic Engineers (IEEE), and Fellow of the Society of Industrial and Applied Mathematics (SIAM).

Title of Talk

The Promise of Machine Learning for Infrared Spectroscopic Image Analysis

A key theme in near and mid-infrared imaging data sciences is the image acquisition coupled to robust computational analysis of spatial and spectral information, to simultaneously predict shape and functional properties of complex scenes. Current capabilities allow for the rapid recording of full hyper-spectral image sets, typically 100 GB, having 10 megapixels, with each pixel containing 2000 spectral wavelengths, with an absorbance value between [0, 1], with noise levels from [10−3, 0.1]. The barriers limiting progress today then is accurate and rapid computational image analysis for the robust elucidation of spatial structure and their material properties.. This talk shall dwell on the success, and current challenges of state of the art machine learning algorithms (spectral de-noising, multi-label classification, functional un-mixing, ..). We quantify the improvement, or lack thereof, using error estimates for image classification and spectral signature recovery.

Prof. Giri Narasimhan
Bioinformatics Research Group (BioRG),
Florida International University
Brief Speaker Bio:
Giri Narasimhan a professor in the School of Computing and Information Sciences at Florida International University (FIU), where he heads the Bioinformatics Research Group (BioRG). He holds a Bachelor’s degree in Electrical Engineering from the Indian Institute of Technology-Bombay and a PhD in Computer Science from the University of Wisconsin-Madison. His interdisciplinary research involves problems from the fields of Algorithms, Bioinformatics, Biotechnology, Data Mining, Machine Learning, Theoretical Computer Science, Graph Theory, Optimization, High Performance Computing, Computational Finance and Statistics. His research has been funded by the National Science Foundation, the National Institutes of Health, National Institute of Justice, Florida Department of Health, Department of Defense and private industry. He has published one monograph, two edited volumes, and over 140 refereed articles in books, journals, and conference proceedings. He is the coach of the FIU Programming Team and is a Co-Director of the Ultimate Software Academy for Teaching and outreach at FIU. As the Associate Dean for Research and Graduate Studies in the College of Engineering and Computing during 2009-15 he managed the research portfolio of the college and oversaw the graduate programs in the college.

Research Group:

Title of Talk

War Stories from the Bioinformatics Front

Widely accepted criteria routinely misdiagnose autoimmune diseases such as Systemic Lupus Erythematosus (SLE) and Mixed Connective Tissue Disease (MCTD). To make matters worse, these criteria fail to differentiate between these two similar diseases. We show how to achieve better accuracies on this front.
Glimpses into microbial communities reveal a diverse, dynamic and complex environment where the entities compete for nutrients, generously share functional genes, produce weapons of destruction in the form of toxins, release various metabolites and signaling molecules for sharing and communication, combine forces to fight common enemies, and much more. We discuss challenges in modeling microbiomes as predictive social networks.

Qing Wu
Brief Speaker Bio:
Qing Wu received the B.S. and M.S. degrees from the Department of Information Science and Electronic Engineering, Zhejiang University, Hangzhou, China, in 1993 and 1995, respectively, and the Ph.D. degree from the Department of Electrical Engineering, University of Southern California, Los Angeles, CA, USA, in 2002. He was an Assistant Professor with the Department of Electrical and Computer Engineering, State University of New York at Binghamton, Binghamton, NY, USA. He is currently a Principal Electronics Engineer with the United States Air Force Research Laboratory, Information Directorate, Rome, NY, USA. He has authored or co-authored over ninety research papers in international journals and conferences. His current research interests include neuromorphic computing architectures, high-performance computing architectures, deep neural networks, and memristor-based neuromorphic circuits and systems.

Title of Talk

An Energy-Efficient Embedded Implementation For Target Recognition In SAR Imageries

We present an energy-efficient deep learning model design, training and implementation method for the synthetic aperture radar (SAR) image classification application on a neuromorphic processor. The proposed approach adopts emerging neuromorphic computing models and hardware to achieve significant improvement in computational energy efficiency over deep learning algorithms on conventional embedded processors. A deep convolutional neural network (DCNN) is designed specifically for implementing image classification on the TrueNorth neurosynaptic processor. We have explored the DCNN model design parameters to obtain a comprehensive solution set in the energy-performance trade-off space. Using a SAR image classification dataset, evaluation results show that the proposed design and implementation approach achieves at least 20X reduction in energy-per-image-classification over one of today’s most energy-efficient conventional embedded processors. while achieving a classification accuracy of 95% and a processing throughput of 1,000 images per second.

Srinivasan Parthasarathy
The Ohio State University
Brief Speaker Bio:
Srinivasan Parthasarathy is a Professor of Computer Science and Engineering and the director of the data mining research laboratory at Ohio State. His research interests span databases, data mining and high performance computing. He is among a handful of researchers nationwide (USA) to have won both the Department of Energy and National Science Foundation Career awards. He and his students have won multiple best paper awards or "best of" nominations from leading forums in the field including: ACM SIGKDD, VLDB, ISMB, WWW, SIAM Data Mining, ICDM, and ACM Bioinformatics. He chairs the SIAM data mining conference steering committee and serves on the action board of ACM TKDD and ACM DMKD --leading journals in the field. Since 2012 he also helped lead the creation of OSU's first-of-a-kind nationwide undergraduate major in data analytics and serves as one of its founding directors.

Title of Talk

Scalable Data Analytics: The Role of Stratified Data Sharding

With the increasing popularity of structured data stores, social networks and Web 2.0 and 3.0 applications, complex data formats, such as trees and graphs, are becoming ubiquitous. Managing and processing such large and complex data stores, on modern computational eco-systems, to realize actionable information efficiently, is daunting. In this talk I will begin with discussing some of these challenges. Subsequently I will discuss a critical element at the heart of this challenge relates to the sharding, placement, storage and access of such tera- and peta- scale data. In this work we develop a novel distributed framework to ease the burden on the programmer and propose an agile and intelligent placement service layer as a flexible yet unified means to address this challenge. Central to our framework is the notion of stratification which seeks to initially group structurally (or semantically) similar entities into strata. Subsequently strata are partitioned within this eco-system according to the needs of the application to maximize locality, balance load, minimize data skew or even take into account energy consumption. Results on several real-world applications validate the efficacy and efficiency of our approach. (Notes: Joint work with Y. Wang (Airbnb) and A. Chakrabarti (MSR).

Sartaj K Sahni
Computer and Information Sciences and Engineering
University of Florida, USA
Brief Speaker Bio :
Sartaj Sahni is a Distinguished Professor of Computer and Information Sciences and Engineering at the University of Florida. He is also a member of the European Academy of Sciences, a Fellow of IEEE, ACM, AAAS, and Minnesota Supercomputer Institute, and a Distinguished Alumnus of the Indian Institute of Technology, Kanpur. In 1997, he was awarded the IEEE Computer Society Taylor L. Booth Education Award ``for contributions to Computer Science and Engineering education in the areas of data structures, algorithms, and parallel algorithms'', and in 2003, he was awarded the IEEE Computer Society W. Wallace McDowell Award ``for contributions to the theory of NP- hard and NP-complete problems''. Dr. Sahni was awarded the 2003 ACM Karl Karlstrom Outstanding Educator Award for ``outstanding contributions to computing education through inspired teaching, development of courses and curricula for distance education, contributions to professional societies, and authoring significant textbooks in several areas including discrete mathematics, data structures, algorithms, and parallel and distributed computing.'' Dr. Sahni has published over three hundred research papers and written 15 texts. His research publications are on the design and analysis of efficient algorithms, parallel computing, interconnection networks, design automation, and medical algorithms. He is presently the Editor-in- Chief of ACM Computing Surveys.

Title of Talk

Time & Energy Efficient Computing

Traditionally, the performance of an application has been measured by its time and memory requirements. More recently, there has been interest in a third performance metric—energy. This talk will begin by motivating the consideration of energy as a metric and then focus on two areas—data centers and big data computations—where both time to completion and energy consumption are important metrics. For data centers, we review some of our recent work in optimizing data center network topology in support of long running data aggregations and, for big data computations, we demonstrate significant time and energy improvements for large scale sequence alignment and RNA folding by enhancing the cache utilization of classical algorithms.

Sanjay Ranka, Ph.D.
Professor, Fellow IEEE, Fellow AAAS
Department of Computer Science and Engineering
University of Florida
Brief Speaker Bio :
Sanjay Ranka is a Professor in the Department of Computer Science. His current research interests are data mining, informatics and grid computing for data intensive applications in High Energy Physics, BioTerrorism and BioMedical Computing. Most recently he was the Chief Technology Officer at Paramark where he developed real-time optimization software for optimizing marketing campaigns. Sanjay has also held positions as a tenured faculty positions at Syracuse University and as a researcher/visitor at IBM T.J. Watson Research Labs and Hitachi America Limited. Sanjay earned his Ph.D. (Computer Science) from the University of Minnesota in 1988 and a B. Tech. in Computer Science from IIT, Kanpur, India in 1985. He has coauthored two books: Elements of Neural Networks (MIT Press) and Hypercube Algorithms (Springer Verlag), 50+ journal articles and 80+ refereed conference articles. He serves on the editorial board of the Journal of Parallel and Distributed Computing and was a past member of the Parallel Compiler Runtime Consortium and the Message Passing Initiative Standards Committee. He was one of the main architects of the Syracuse Fortran 90D/HPF compiler. He is a fellow of the IEEE and AAAS, advisory board member of IEEE Technical Committee on Parallel Processing and a member of IFIP Committee on System Modeling and Optimization.

Title of Talk

Smart Intersection Control Algorithms for Automated and Connected Vehicles

The pace at which autonomous and connected vehicle technology is reaching the consumer market is accelerating. These driver and driverless vehicles will communicate information with other vehicles and transportation infrastructure. In this talk, we will present our work on optimizing signal control to take advantage of these developments. We will provide extensive simulation results as well as our initial testing of our algorithms On Traffic Engineering Research Laboratory (TERL) in Tallahassee, Florida.
We will also describe related projects on I-STREET (Implementing Solutions from Transportation Research and Evaluation of Emerging Technologies) that University of Florida is developing in conjunction with the Florida Department of Transportation and the City of Gainesville. This testbed will encompass the University of Florida campus and the surrounding roadway network in the City of Gainesville.