Zhuoxun Yang, a first-year student in the Computer Science department, along with Longtao Zhang and Ruoyu Li, second-year students, have recently contributed to advancing scientific data compression through their paper titled “IPComp: Interpolation Based Progressive Lossy Compression for Scientific Applications.” This work has been accepted to the 34th ACM International Symposium on High-Performance Parallel and Distributed Computing (HPDC’25), held in Notre Dame, IN, USA, from July 20–23, 2025, and was selected as a Best Paper Nominee. The project was conducted under the guidance of Dr. Kai Zhao, in collaboration with researchers from Argonne National Laboratory.
Scientific simulations and experiments are producing data at an unprecedented scale, presenting major challenges in storage, transmission, and analysis. Lossy compression has become essential for managing these large datasets, but many existing techniques face trade-offs between compression ratio, accuracy, and computational efficiency.
In this work, the team introduces IPComp, a novel Interpolation-based Progressive lossy compression framework tailored for scientific data. IPComp employs adaptive interpolation to model scientific datasets and incorporates a progressive encoding mechanism that supports multi-resolution reconstruction. This enables efficient storage with controllable accuracy, making it particularly suitable for large-scale scientific applications. Evaluations across real-world datasets demonstrate that IPComp achieves superior compression performance, fidelity, and scalability compared to state-of-the-art methods.
This recognition at HPDC’25 highlights the group’s growing impact on high-performance computing and data-intensive scientific research.