The first deep-learning search for radio technosignatures from 820 nearby stars

The goal of the Search for Extraterrestrial Intelligence (SETI) is to quantify the prevalence of technological life beyond Earth via their ``technosignatures". One theorized technosignature is narrowband Doppler drifting radio signals. The principal challenge in conducting SETI in the radio domain is developing a generalized technique to reject human radio frequency interference (RFI) that dominates the features across the band in searches for technosignatures. Here, we will present the first comprehensive deep-learning-based technosignature search to date, returning 8 promising ETI signals-of-interest for re-observation as part of the Breakthrough Listen initiative. The search comprises 820 unique targets observed with the Robert C. Byrd Green Bank Telescope, totaling over 480hr of on-sky data. We implement a novel beta-Convolutional Variational Autoencoder with an embedded discriminator combined with Random Forest Decision Trees to classify technosignature signals of interest in a semi-unsupervised manner. We compare our results with prior classical techniques on the same dataset and conclude that our algorithm returns more convincing signals of interest with a manageable false positive rate. This novel approach presents itself as a leading solution in accelerating SETI and other transient research into the new age of data-driven astronomy.