Temporal Convolutional Networks for Exoplanet Transit Timing Variations

Transit timing variations (TTVs) in exoplanetary systems can help reveal hidden planets and planetary dynamics. This paper introduces a machine learning approach using Temporal Convolutional Networks (TCNs) to detect and analyze TTVs. We develop a TCN model, trained on transit and nontransit light curves, to capture complex temporal patterns. Our method is validated using data from known multi-planet systems, demonstrating high accuracy in TTV detection. When applied to a dataset of light curves from NASA telescopes Kepler (including K2) and TESS, this network identifies new TTV candidates, helping discover previously undetected planets.