06/15/2021
By Sokny Long

The Francis College of Engineering, Department of Electrical and Computer Engineering, invites you to attend a doctoral dissertation defense by Jialing Zhang on “Data Reduction with Transform based Lossy Compression for Scientific Simulation Frameworks.”

PhD Candidate: Jialing Zhang
Defense Date: Tuesday, June 29, 2021
Time: 9 to 11 a.m.
Location: This will be a virtual defense via Zoom. Those interested in attending should contact jialing_zhang@student.uml.edu and committee advisor, seungwoo_son@uml.edu, at least 24 hours prior to the defense to request access to the meeting.

Committee Chair (Advisor): Seung Woo Son, Associate Professor, Electrical and Computer Engineering, University of Massachusetts Lowell

Committee Members:

  • Martin Margala, Professor, Electrical and Computer Engineering, University of Massachusetts Lowell
  • Hengyong Yu, Professor, Electrical and Computer Engineering, University of Massachusetts Lowell
  • Joyita Dutta, Associate Professor, Electrical and Computer Engineering, University of Massachusetts Lowell

Brief Abstract:
High-performance computing (HPC) applications produce extreme volumes of data that capture simulation states for failure/restart and post-simulation data analysis. Transferring and archiving these raw data generated on supercomputers result in a massive burden on I/O and storage systems. To alleviate this problem, one approach is to reduce the amount of data through compression techniques before storing to disk or transferring through network. This dissertation focuses on developing an efficient and effective data reduction technique for scientific simulation frameworks. This is achieved by applying the idea of preserving dominant data features with the least amount of bits possible. The major challenge in this work is not only to achieve high compression ratios as the simulations generate data far more rapidly than the current scientific workflow can handle, but also to provide high precision and viable decompressed data for simulation frameworks. This thesis investigates and addresses the difficulties in (1) analyzing information retrieval methods and their feature preservation properties, (2) finding the best trade-off solution between information loss and compression ratio, and (3) optimizing the quantization model and encoding mechanism. The technique developed in this work demonstrates superior compression ratios at medium to high compression accuracy compared with state-of-the-art lossy compressors on most of the evaluated scientific datasets.

All interested students and faculty members are invited to attend the online defense via remote access.