CC-AOS Optimizes Time-Series Optimal Stopping Decisions

Finite-horizon optimal stopping is a critical problem in early time-series classification, where a system must decide at each sequence prefix whether the expected benefit of another observation justifies its acquisition cost. Existing data-driven backward-induction methods typically solve each cost-horizon operating point separately, leading to inefficiencies as changing operating conditions require repeated optimization and separate model stacks. Researchers propose CC-AOS (Cost- and Horizon-Conditioned Amortized Optimal Stopping), a structured amortized solver designed for a family of finite-horizon stopping problems with continuous costs and multiple horizons. CC-AOS learns a single, shared continuation-value model that is dynamically conditioned on the current state, absolute time, remaining horizon, and the acquisition cost. This is achieved through joint amortized fitted backward induction. The framework incorporates established properties of exact value and continuation functions, such as being nondecreasing, concave, and horizon-dependently Lipschitz in cost, directly into the model architecture. Experiments on controlled Gaussian and time-varying non-Gaussian processes, as well as the FordA engine-noise time-series benchmark, demonstrated that a single CC-AOS checkpoint consistently achieved a lower terminal-risk-plus-sampling-cost objective than independently fitted Convex Function Learning solvers across six unseen FordA cost-horizon pairs, with an average reduction of 15.75 percent, while matching tuned static thresholds on average.