Workshops Archive

Machine Learning for a Gamma-Ray Telescope Using Stream-HLS


Workshop: Twelfth International Workshop on Heterogeneous High-Performance Reconfigurable Computing (H2RC 2026)

Authors: Jackie Wu and Ye Htet (Washington University in St. Louis), Marion Sudvarg (Missouri University of Science & Technology), and Jeremy D. Buhler and Roger D. Chamberlain (Washington University in St. Louis)

Abstract: Machine learning is becoming an important contributor to data acquisition for science, performing many tasks more efficiently than previously utilized techniques. Scientific instrumentation frequently is required to separate the actual signals of interest from background signals, and machine learning has often been applied to this task. However, manually designing and tuning the HLS implementation of onboard neural network inference is labor-intensive, and the resulting designs often have significant room for improvement.

Automated dataflow HLS tools could be used to implement inference on an instrument's onboard FPGA. We focus on a recent tool - Stream-HLS. Here, we assess the suitability of the Stream-HLS framework to implement a background rejection task for a gamma-ray telescope. The Stream-HLS design proves to be able to fit the flight part (XC7K325T), whereas the hand-tuned baseline does not. Moreover, the Stream-HLS design achieves ~1.9X higher throughput, while leaving 43% of the available DSPs idle.


Back to Twelfth International Workshop on Heterogeneous High-Performance Reconfigurable Computing (H2RC 2026) Archive Listing

Back to Full Workshop Archive Listing