Modeling Dependence Structures in Astronomical Multi-Band Time Series Data via Multi-Output Gaussian Processes
The study examines three ways to model how different photometric bands vary together in astronomical time series using multi-output Gaussian processes: covariance-based, SDE-based, and latent-process approaches. The key finding is that the choice of dependence structure significantly impacts scientific interpretation; for instance, one approach is better for characterizing observed variability while another is superior for inferring physical mechanisms like reverberation mapping.