doi: 10.18178/joebm.2026.14.4.964
Learning Optimal Stopping Policies via High-Precision Teacher Labels
Lanqiong Lou
- Department of Applied Mathematics and Statistics, Johns Hopkins University, Baltimore, USA
- Manuscript received July 21, 2026
- revised October 12, 2026
- published October 27, 2026
Abstract
This paper discusses the effect of labeling on classification-based optimal stopping. Labels generated naively from one continuation payoff realization contain a lot of pathwise noise since the label does not capture the conditional expectation of continuation. In order to resolve this problem, a High-Precision LSM Teacher generates labels for exercises or continues while a quadratic Logistic Regression Student learns from these labels. Comparing labels directly, it turns out that using Teacher labels instead of pathwise labels decreases the pricing error of the Student relative to the Teacher from 16.09% to 0.51%. Using SVMs for comparison, there appear no significant improvements in using more complicated classifiers.
Keywords
- American option pricing
- optimal stopping
- least-squares monte Carlo
- teacher–student learning
- logistic regression
How to Cite
Lanqiong Lou, "Learning Optimal Stopping Policies via High-Precision Teacher Labels," Journal of Economics, Business and Management, vol. 14, no. 4, pp. 324-339, 2026. https://doi.org/10.18178/joebm.2026.14.4.964
Copyright & License
Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).