Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.18751 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908795040432128 |
|---|---|
| author | Yoon, Samuel Kim, Jongwon Ha, Juyoung Ko, Young Myoung |
| author_facet | Yoon, Samuel Kim, Jongwon Ha, Juyoung Ko, Young Myoung |
| contents | Recently reconstruction-based deep models have been widely used for time series anomaly detection, but as their capacity and generalization capability increase, these models tend to over-generalize, often reconstructing unseen anomalies accurately. Prior works have attempted to mitigate this by incorporating a memory architecture that stores prototypes of normal patterns. Nevertheless, these approaches suffer from high training costs and have yet to be effectively integrated with time series foundation models (TFMs). To address these challenges, we propose MOMEMTO, an improved variant of TFM for anomaly detection, enhanced with a patch-based memory module to mitigate over-generalization. The memory module is designed to capture representative normal patterns from multiple domains and enables a single model to be jointly fine-tuned across multiple datasets through a multi-domain training strategy. MOMEMTO initializes memory items with latent representations from a pre-trained encoder, organizes them into patch-level units, and updates them via an attention mechanism. We evaluate our method using 23 univariate benchmark datasets. Experimental results demonstrate that MOMEMTO, as a single model, achieves higher scores on AUC and VUS metrics compared to baseline methods, and further enhances the performance of its backbone TFM, particularly in few-shot learning scenarios. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_18751 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | MOMEMTO: Patch-based Memory Gate Model in Time Series Foundation Model Yoon, Samuel Kim, Jongwon Ha, Juyoung Ko, Young Myoung Machine Learning Recently reconstruction-based deep models have been widely used for time series anomaly detection, but as their capacity and generalization capability increase, these models tend to over-generalize, often reconstructing unseen anomalies accurately. Prior works have attempted to mitigate this by incorporating a memory architecture that stores prototypes of normal patterns. Nevertheless, these approaches suffer from high training costs and have yet to be effectively integrated with time series foundation models (TFMs). To address these challenges, we propose MOMEMTO, an improved variant of TFM for anomaly detection, enhanced with a patch-based memory module to mitigate over-generalization. The memory module is designed to capture representative normal patterns from multiple domains and enables a single model to be jointly fine-tuned across multiple datasets through a multi-domain training strategy. MOMEMTO initializes memory items with latent representations from a pre-trained encoder, organizes them into patch-level units, and updates them via an attention mechanism. We evaluate our method using 23 univariate benchmark datasets. Experimental results demonstrate that MOMEMTO, as a single model, achieves higher scores on AUC and VUS metrics compared to baseline methods, and further enhances the performance of its backbone TFM, particularly in few-shot learning scenarios. |
| title | MOMEMTO: Patch-based Memory Gate Model in Time Series Foundation Model |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2509.18751 |