Case Study: NCSA Gravity Group achieves real-time gravitational wave analysis with NVIDIA GPUs

Seeing Gravity In Real-time With Deep Learning

NCSA Gravity Group at the University of Illinois at Urbana-Champaign studies astrophysics and gravitational waves, but detecting and characterizing these signals in highly noisy data is computationally demanding and must be done with very low latency to support real-time, multi-messenger astronomy. NVIDIA helped the team address this challenge with NVIDIA Tesla GPUs and the MXNet framework.

NVIDIA and the NCSA Gravity Group developed and trained a 15-layer deep convolutional neural network to classify weak gravitational-wave signals and predict black hole masses from raw time-series data. The solution improved AI inference performance by 100x, with NVIDIA GPUs adding another 50x boost, delivering more than a 1,000x overall gain and enabling real-time analysis that supports faster follow-up observations and the discovery of previously missed gravitational-wave sources.

NCSA Gravity Group

Eliu Huerta

Head of the NCSA Gravity Group