A new sampler for the Laser Interferometer Space Antenna
The samsara++ sampler helps disentangle overlapping gravitational-wave signals.
Researchers developed samsara++, a C++/CUDA sampler. It is designed for the Laser Interferometer Space Antenna (LISA).
LISA will observe a superposition of gravitational-wave signals. These signals overlap in both time and frequency. A Global Fit (GF) analysis is needed to disentangle these contributions.
The algorithm uses birth-death processes for adding and removing sources. It also uses mutation moves for parameter estimation. Birth and death are selected through posterior-dependent rates.
Tests used simulated one-year LISA datasets. These datasets contained Galactic double white dwarf binaries and stationary Gaussian instrumental noise. One test used 143 and 866 binaries.
A stress test explored a population of 16,435 binaries. This involved 131,480 source parameters. The run completed 15 million iterations in approximately 16 hours using about 3 GB of RAM.
The analysis highlights the importance of source-count, overlapping signals, and incomplete mixing. These factors affect joint source and noise inference.
Key facts
- samsara++ uses birth-death processes to add or remove sources.
- The sampler completed 15 million iterations in about 16 hours.
- The tool was tested on populations of 143 and 866 binaries.
- The stress test used approximately 3 GB of RAM.