Modeling Doppler Shifts in Radial-Velocity Data with Deep Learning toward Earth-mass Exoplanet Detection
Researchers have developed a deep learning framework to improve the detection of Earth-mass exoplanets by analyzing Doppler shifts in radial-velocity data. This new model, trained on HARPS-N solar spectra, utilizes physics-motivated spectral representations and genetic algorithms for hyperparameter optimization. The framework successfully identifies planetary signals with amplitudes as low as 25 cm/s and periods between 10 and 550 days, outperforming traditional flux-based methods. The team has also released a Python package called doppleriann to implement this framework, offering a promising approach for detecting exoplanets in real observational data. AI
IMPACT This research demonstrates a novel application of deep learning in astrophysics, potentially accelerating the discovery of Earth-like exoplanets.