Friday, February 25

MS64
Parallel Processing in Data Science Applications - Part I of II

3:35 PM - 5:15 PM

For Part II, see MS65

Some of the most challenging problems in data-driven science involve understanding the interactions between millions or even thousands of millions of variables. In the era of Big Data, we are faced with the ever-increasing size, variability, and uncertainty of the datasets which are challenging high-performance numerical methods and software for extreme-scale computing. Among these data-intensive applications are problems arising from finance, biomedical applications, social networks, image classification or climate data. In recent years, deep neural networks have become invaluable for machine learning based classification algorithms. Besides, parallel processing algorithms such as hypergraph partitioning methods or algorithms exploiting the structure of the associated graph Laplacian as well as the Bayesian approach are further computational key methods for big data analytics. The vast quantity and variety of big data applications leads to bottlenecks in storage and computational costs while at the same time green computing becomes indispensable to reduce the carbon footprint even on distributed platforms such as clusters, grids, and clouds.

Organizer: Matthias Bollhöfer
Technische Universität Braunschweig, Germany
Olaf Schenk
Università della Svizzera italiana, Switzerland

3:35-3:55 Should You Go with the Flow ? A New Tensor Algebra for Neural Networks abstract

Lior Horesh, IBM Research, U.S.; Misha E. Kilmer, Tufts University, U.S.; Elizabeth Newman, Emory University, U.S.; Shashanka Ubaru, IBM T.J. Watson Research Center, U.S.; Osman Malik, University of Colorado Boulder, U.S.; Haim Avron, Tel Aviv University, Israel

4:00-4:20 Handling Runtime Variability at Scale: Elasticity for Continuously Changing Graphs abstract

Kasimir Gabert, Georgia Institute of Technology, U.S.; Ali Pinar, Sandia National Laboratories, U.S.; Umit V. Catalyurek, Georgia Institute of Technology, U.S.

4:25-4:45 Sparse Precision Matrix Estimation for Large-Scale Datasets abstract

Aryan Eftekhari, Università della Svizzera italiana, Switzerland ; Matthias Bollhöfer, Technische Universität Braunschweig, Germany; Olaf Schenk, Università della Svizzera italiana, Switzerland

4:50-5:10 Multi-Scale Methods for Machine Learning abstract

Inderjit S. Dhillon, University of Texas at Austin, U.S.