Speaker
Luca Negri
(Utrecht University, Nikhef)
Description
The computational costs of gravitational wave inference is expected to exponentially rise with the next generation of detectors: both the complexity and the amount of data itself will be much higher, requiring a complete rethinking of current parameter estimation methods to produce accurate science without prohibitive resources usage.
This work will present a novel way of dramatically reducing the computational costs of Markov-chain-monte-Carlo algorithms by approximating the analytical Bayesian likelihood with a Neural likelihood estimator. This method obtains compatible posteriors and returns the correct Bayesian evidence, requiring only a fraction of waveforms computations compared to standard methods.
| AI keywords | Simulation-based inferece; parameter estimation; bayesian statistichs; speed-up |
|---|
Primary author
Luca Negri
(Utrecht University, Nikhef)
