AI and the World's Hard Problems · Astronomy & space · arXiv · 2026-10-20
Single-Transit Parameter Estimation with Calibrated Uncertainties
Single transit events are hard to analyze with standard methods designed for periodic signals, yet characterizing them matters for follow-up. The authors trained a convolutional neural network on synthetic light curves to predict transit depth, duration, and mid-transit time with calibrated uncertainties from one transit. On 120 confirmed exoplanets it matched a trapezoidal least-squares baseline in median error while failing in 1% of cases versus 10% for the baseline, and it identified 169 new candidates in TESS Sectors 27–60.