Uploaded July 2022 | Updated September 2026, 2 hours ago
Massive Black Hole Binaries (MBHBs) are one of the main astrophysical sources that we expect to detect in the LISA band. Despite their potentially very high Signal to Noise Ratio (SNR), estimating the parameters that describe the source is a non-trivial problem.
State-of-the-art methods perform this task by sampling from the posterior sample, which, due to their highly-dimensional, multimodal posteriors, require very large amounts of simulations to be run. This makes data analysis a slow, computationally expensive process to perform every time a new candidate source is detected.
Machine learning methods may be able to pay a comparatively similar or smaller cost in upfront training, but should be very quick once that training is finished. We will be exploring these types of methods as applied to the task of MBHB parameter estimation and describe the progress of our research in this front.
Authors:
Ivan Martin Vilchez (Presenting)
Institute of Space Sciences (ICE, CSIC), Cerdanyola del Vallès (Barcelona), Spain
Institute of Space Studies of Catalonia (IEEC), Barcelona, Spain
Universitat Autònoma de Barcelona (UAB), Bellaterra (Barcelona), Spain
Carlos Sopuerta
Institute of Space Sciences (ICE, CSIC), Cerdanyola del Vallès (Barcelona), Spain
Institute of Space Studies of Catalonia (IEEC), Barcelona, Spain
Massive Black Hole Binaries (MBHBs) are one of the main astrophysical sources that we expect to detect in the LISA band. Despite their potentially very high Signal to Noise Ratio (SNR), estimating the parameters that describe the source is a non-trivial problem.
State-of-the-art methods perform this task by sampling from the posterior sample, which, due to their highly-dimensional, multimodal posteriors, require very large amounts of simulations to be run. This makes data analysis a slow, computationally expensive process to perform every time a new candidate source is detected.
Machine learning methods may be able to pay a comparatively similar or smaller cost in upfront training, but should be very quick once that training is finished. We will be exploring these types of methods as applied to the task of MBHB parameter estimation and describe the progress of our research in this front.
Authors:
Ivan Martin Vilchez (Presenting)
Institute of Space Sciences (ICE, CSIC), Cerdanyola del Vallès (Barcelona), Spain
Institute of Space Studies of Catalonia (IEEC), Barcelona, Spain
Universitat Autònoma de Barcelona (UAB), Bellaterra (Barcelona), Spain
Carlos Sopuerta
Institute of Space Sciences (ICE, CSIC), Cerdanyola del Vallès (Barcelona), Spain
Institute of Space Studies of Catalonia (IEEC), Barcelona, Spain
