early-time-series-classification

This paper proposes CC-AOS, a structured amortized solver for finite-horizon optimal stopping problems that handles varying costs and horizons without retraining. It incorporates theoretical properties into the model architecture and demonstrates improved performance on benchmark tasks.