Nolan Smyth

UC Santa Cruz

“Learning Pulsar Timing Data”

This work applies machine learning to the task of detecting a stochastic gravitational wave background. This allows us to efficiently analyze a large data set, looking for correlations in the arrival times of light pulses from rapidly rotating neutron stars.


Tracking the time of arrivals of light pulses from rapidly rotating pulsars is the leading method to detect gravitational waves (GW) in the nanoHertz frequency range generated by merging supermassive black hole binaries or primordial fluctuations. As the size of the relevant data set continues to grow with the discovery of more pulsars and the larger number of observations, existing analysis methods become increasingly inefficient and slow. In this work, we introduce machine learning techniques to the pulsar timing array analysis. By front loading the bulk of the computation into training a neural network, we create a near instantaneous analysis pipeline, allowing highly efficient parameter estimation.

12 + 10 =