Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.11367 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914036458717184 |
|---|---|
| author | Lee, Chang-Hwan Shim, Alexander |
| author_facet | Lee, Chang-Hwan Shim, Alexander |
| contents | Reinforcement learning (RL) agents typically assume stationary environment dynamics. Yet in real-world applications such as healthcare, robotics, and finance, transition probabilities or reward functions may evolve, leading to model drift. This paper proposes a novel framework to detect such drifts by analyzing the distributional changes in sequences of agent behavior. Specifically, we introduce a suite of edit operation-based measures to quantify deviations between state-action trajectories generated under stationary and perturbed conditions. Our experiments demonstrate that these measures can effectively distinguish drifted from non-drifted scenarios, even under varying levels of noise, providing a practical tool for drift detection in non-stationary RL environments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_11367 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Detecting Model Drifts in Non-Stationary Environment Using Edit Operation Measures Lee, Chang-Hwan Shim, Alexander Machine Learning Artificial Intelligence Reinforcement learning (RL) agents typically assume stationary environment dynamics. Yet in real-world applications such as healthcare, robotics, and finance, transition probabilities or reward functions may evolve, leading to model drift. This paper proposes a novel framework to detect such drifts by analyzing the distributional changes in sequences of agent behavior. Specifically, we introduce a suite of edit operation-based measures to quantify deviations between state-action trajectories generated under stationary and perturbed conditions. Our experiments demonstrate that these measures can effectively distinguish drifted from non-drifted scenarios, even under varying levels of noise, providing a practical tool for drift detection in non-stationary RL environments. |
| title | Detecting Model Drifts in Non-Stationary Environment Using Edit Operation Measures |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2509.11367 |