Global Fit Inference for LISA with samsara++

The Laser Interferometer Space Antenna (LISA) will observe a superposition of gravitational-wave signals, requiring joint inference of an unknown number of sources, parameters, and instrumental noise. To disentangle these contributions, that overlap in both time and frequency, a Global Fit (GF) analysis is needed. We present samsara++, a C++/CUDA sampler designed to address this trans-dimensional GF problem. Building on the continuous-time MCMC framework of samsara, the algorithm combines birth-death processes for adding and removing sources, with mutation moves for parameter estimation. Birth and death are selected through posterior-dependent rates. Shared-memory parallelism, incremental likelihood updates, and indexed storage reduce the computational and memory costs, while parallel tempering improves the exploration of multimodal posteriors. We assess the performance using simulated one-year LISA datasets containing Galactic double white dwarf binaries and stationary Gaussian instrumental noise. Analyses of populations containing 143 and 866 binaries start from empty source models, without preliminary searches or proposals tailored to the injected population. These tests demonstrate source reconstruction and catalog generation, highlighting, at the same time, the importance of source-count, overlapping signals, and incomplete mixing on joint source and noise inference. Then we perform a separate stress test, initialized at the injected parameters, exploring a population of 16,435 binaries, corresponding to 131,480 source parameters. This run completes 15 million iterations in approximately 16 hours using about 3 GB of RAM. All together, these benchmarks show the computational feasibility of birth-death sampling for large Galactic-binary populations, and provide a foundation for extending samsara++ toward a comprehensive LISA global fit.

Authors

Publication Details

Journal
arXiv (Cornell University)
Published
2026-11-12
DOI
https://doi.org/10.48550/arxiv.2611.17266
Primary Topic
Pulsars and Gravitational Waves Research
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Global Fit Inference for LISA with samsara++

Angelo Ricciardone, R. Buscicchio, Walter da Pozzo, Lorenzo Valbusa Dall'Armi et al.
arXiv (Cornell University)
Pulsars and Gravitational Waves Research
preprint

Global Fit Inference for LISA with samsara++

Angelo Ricciardone, R. Buscicchio, Walter da Pozzo, Lorenzo Valbusa Dall'Armi, Gabriele Astorino
preprint en

Abstract

The Laser Interferometer Space Antenna (LISA) will observe a superposition of gravitational-wave signals, requiring joint inference of an unknown number of sources, parameters, and instrumental noise. To disentangle these contributions, that overlap in both time and frequency, a Global Fit (GF) analysis is needed. We present samsara++, a C++/CUDA sampler designed to address this trans-dimensional GF problem. Building on the continuous-time MCMC framework of samsara, the algorithm combines birth-death processes for adding and removing sources, with mutation moves for parameter estimation. Birth and death are selected through posterior-dependent rates. Shared-memory parallelism, incremental likelihood updates, and indexed storage reduce the computational and memory costs, while parallel tempering improves the exploration of multimodal posteriors. We assess the performance using simulated one-year LISA datasets containing Galactic double white dwarf binaries and stationary Gaussian instrumental noise. Analyses of populations containing 143 and 866 binaries start from empty source models, without preliminary searches or proposals tailored to the injected population. These tests demonstrate source reconstruction and catalog generation, highlighting, at the same time, the importance of source-count, overlapping signals, and incomplete mixing on joint source and noise inference. Then we perform a separate stress test, initialized at the injected parameters, exploring a population of 16,435 binaries, corresponding to 131,480 source parameters. This run completes 15 million iterations in approximately 16 hours using about 3 GB of RAM. All together, these benchmarks show the computational feasibility of birth-death sampling for large Galactic-binary populations, and provide a foundation for extending samsara++ toward a comprehensive LISA global fit.

arXiv (Cornell University)
Pulsars and Gravitational Waves Research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon