LIGO Document T2400274-v1
- Gravitational-wave (GW) science has opened new avenues for understanding
astrophysical phenomena, with precise signal characterization being essential
for interpreting these cosmic events. However, short-duration terrestrial noise
transients, known as “glitchesâ€, complicate this task. A common strategy for
mitigating the impact of glitches involves restricting the analysis to a reduced
frequency band. In my work, I utilize a neural posterior estimator, a deep
learning (DL) model, to investigate how varying frequency bands influence GW
parameter estimation for compact binary coalescences. I perform parameter
estimation on artificial data, which include both simulated waveforms and noise
realizations. By examining different frequency ranges, I aim to characterize the
typical effects of these restrictions on the inference of the source parameters.
Additionally, I focus on evaluating the use of a DL model in the GW parameter
estimation pipeline. By comparing results obtained with the DL model and the
conventional pipeline, I explore the effectiveness of deep learning approaches in
improving analysis speed and accuracy. My ongoing work involves refining the
network’s capabilities and analyzing the impact of varying frequency bands on
the reliability of parameter estimation.
DCC Version 3.5.2, contact
DCC Help