Saturday, February 15

MS65
High-Performance Tensor Computation and Applications - Part II of III

10:40 AM - 12:20 PM
Room: 508

For Part I, see MS55

For Part III, see MS74

Tensors are higher order generalization of matrices that provide a natural way to represent a multi-relational dataset. Given a dataset encoded as a tensor, tensor decomposition serves as a promising analytics tool for mining this data to uncover hidden structure within the data's relations.

This minisymposium explores efficient and scalable solutions for calculating tensor decomposition, as well as its application in data analytics across areas spanning signal processing, cybersecurity, machine learning, and beyond.

Organizer: Jee W. Choi
University of Oregon, U.S.
Rich Vuduc
Georgia Institute of Technology, U.S.
Eric Phipps
Sandia National Laboratories, U.S.

10:40-11:00 Scaling Up Streaming Tensor Decompositions abstract

Shaden Smith, Intel, AI, U.S.

Cancelled 11:05-11:25 Hpc_td_tbd_battaglino

Casey Battaglino, Georgia Institute of Technology, U.S.

11:30-11:50 Stochastic Gradients for Large-Scale Tensor Decomposition abstract

Tamara Kolda, Sandia National Laboratories, U.S.; David Hong, University of Pennsylvania, U.S.