Singapore researchers build AI framework for black-hole studies
シンガポールの研究者、ブラックホール研究向けAIを開発科学・技術
- Country:
- 🇸🇬 Singapore
- Voice:
- 🇬🇧 Sophie
英語本文
Researchers at the National University of Singapore have developed an AI framework to infer the properties of orbiting pairs of supermassive black holes from pulsar timing data. The university described the work by assistant professor Alvin Chua and postdoctoral researcher Subhajit Dandapat on October 22. Reported improvements came from simulated tests, not a newly confirmed detection of an individual pair.
Pulsars are rapidly rotating neutron stars whose regular radio pulses serve as cosmic clocks. Passing gravitational waves can cause tiny, correlated changes in the pulses' arrival times. The framework incorporates the physics of black-hole orbital motion, combining Transformer models with simulation-based inference. It trains on millions of simulated pulsar signals.
The model estimates masses, orbital frequency, orbital eccentricity and sky position, together with uncertainty. The team reported its largest improvements when training data were limited. Next, the researchers plan to test more realistic datasets containing different sources of noise and other gravitational-wave signals.