Speaker
Anastasios Theodoropoulos
Description
Gravitational wave approximants are widely used in the analysis of Gravitational wave signals. They are highly accurate but they still lack speed. We present an autoencoder based surrogate model for BBH gravitational wave signals. We train our autoencoder on a NR informed surrogate model, we modify it so it can produce waveforms from the initial binary system parameters and then we fine tune it using the NR waveforms in the SXS dataset. We achieve an accurate approximant, having mismatches of the order of 10-4 with the NR waveforms, while being able to produce 106 waveforms in under a second.
