Sujata Sinha Wins Best Paper Award at HPDC 2026
The Virginia Tech researcher earned top honors in Cleveland for her work on scientific data compression.
Updated on Oct. 26, 2026 in Science — General
In July 2026, PhD candidate Sujata Sinha received the best paper presentation award at the HPDC 2026 symposium held in Cleveland. She presented research on an advanced framework for lossy data compression.
Why it matters
The study addresses shortcomings in classical rate-distortion theory by accounting for the specific heterogeneity and tile-based structures found in modern scientific datasets.
The research team utilized 1D and 2D Gaussian models to test SPERR and SVD compressors. The study specifically analyzes spatial correlations and error-bounded tile size parameters to optimize compression efficiency.
The players
Sujata Sinha
She is a PhD candidate at Virginia Tech and a visiting researcher at Argonne National Laboratory.
Robert Underwood
He served as a supervisor for the research project presented at the symposium.
Lingjia Liu
He provided academic supervision for the development of the compression research.
The details
Sinha's paper, titled 'Bridging Information Theory and Practice for Scientific Lossy Compression,' was supervised by researchers Robert Underwood and Lingjia Liu. The framework successfully balances compression efficiency with the requirements of parallel computing systems.
Timeline
July 2026: Sujata Sinha presented the research paper at the HPDC symposium.
The Big Picture
The research extends the methodologies used in the Argonne National Laboratory high-performance computing research program. This work follows the pattern of data-management innovation currently being developed within that institutional framework.
This research could lead to more efficient data storage solutions for high-performance computing environments. These improvements potentially accelerate the processing of large-scale scientific datasets in research and industry.
The takeaway
Researchers can achieve higher data efficiency by moving beyond classical models to account for spatial correlations. Implementing these specific tiling parameters may help teams optimize their own large-scale data workflows.
Further reading
Learn more about local scientific progress in the Science — General section.
Source note: This article includes information reported by Argonne National Laboratory.
