Scalable Algorithms for Tensor Computations

Speaker: Edgar Solomonik

Date: Wednesday, February 24, 2021

Time: 3:00 PM to 4:00 PM Note: all times are in the Eastern Time Zone

Public: Yes

Location: (Registration required, if you haven't registered for this series before)

Event Type: Seminar

Room Description:

Host: Julian Shun, MIT CSAIL

Contact: Julian Shun,,

Relevant URL:

Speaker URL:

Speaker Photo:

Reminders to:,,,

Reminder Subject: TALK: Scalable Algorithms for Tensor Computations

Abstract: We describe recent advances in algorithms and software for tensor contraction, tensor decomposition, and tensor networks, as well as their application to problems in quantum simulation. These include more efficient formulations of nonlinear optimization methods for tensor decomposition, completion, and tensor networks. We develop parallel implementations of these methods by leveraging the Cyclops library for dense and sparse tensor contractions. We also describe a new automatic differentiation library for tensor code (AutoHOOT). AutoHOOT makes use of tensor algebra identities to generate code for high-order optimization methods that is orders of magnitude faster than that generated by commercial automatic differentiation software such as JAX. Driven by consideration of structure in tensors arising in quantum simulation methods, we also devise more efficient methods for contraction of tensors with permutational or group symmetry.

Bio: Edgar Solomonik has been an Assistant Professor in the Department of Computer Science at the University of Illinois at Urbana-Champaign since 2016. He received his PhD from UC Berkeley in 2014 and was subsequently a postdoctoral fellow at ETH Zurich. He is known for work on tensor computations and communication-avoiding algorithms. His recognitions include the NSF CAREER award, the SIAM Activity Group on Supercomputing Early Career Prize, the IEEE CS TCHPC Award for Excellence for Early Career Researchers in High Performance Computing, and the Householder Prize.

IMPORTANT NOTE FOR ATTENDEES: If you have already registered for the Fast Code Seminars on Zoom since July 27, 2020, please use the Zoom link that you have received. This link will stay the same for subsequent Fast Code seminars this semester. Zoom does not recognize a second registration, and will not send out the link a second time. If you have any problems with registration, please contact and by 1:30pm on the day of the seminar, so that we can try to resolve it before the seminar begins.

Research Areas:
Algorithms & Theory, Programming Languages & Software

Impact Areas:
Big Data

See other events that are part of the Fast Code 2020 - 2021.

Created by Julian J. Shun Email at Thursday, February 18, 2021 at 5:19 PM.