{"paper":{"title":"PaAno: Patch-Based Representation Learning for Time-Series Anomaly Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A patch-based CNN method for time-series anomaly detection surpasses complex models on benchmarks.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Jinju Park, Seokho Kang","submitted_at":"2026-02-01T17:52:59Z","abstract_excerpt":"Although recent studies on time-series anomaly detection have increasingly adopted ever-larger neural network architectures such as transformers and foundation models, they incur high computational costs and memory usage, making them impractical for real-time and resource-constrained scenarios. Moreover, they often fail to demonstrate significant performance gains over simpler methods under rigorous evaluation protocols. In this study, we propose Patch-based representation learning for time-series Anomaly detection (PaAno), a lightweight yet effective method for fast and efficient time-series "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"PaAno achieved state-of-the-art performance, significantly outperforming existing methods, including those based on heavy architectures, on both univariate and multivariate time-series anomaly detection across various range-wise and point-wise performance measures.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the learned patch embeddings reliably separate normal from anomalous temporal patterns and that comparison to training-set normal patches produces a valid anomaly score without additional calibration or post-hoc tuning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"PaAno uses patch-based 1D CNN embeddings trained with triplet and pretext losses to achieve state-of-the-art time-series anomaly detection on the TSB-AD benchmark for both univariate and multivariate data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A patch-based CNN method for time-series anomaly detection surpasses complex models on benchmarks.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"938f1bbcb899f0b49898f6c425239fc628b523019331db1e94fda65fca6cf873"},"source":{"id":"2602.01359","kind":"arxiv","version":2},"verdict":{"id":"884e849a-3998-4b5d-a47c-4574aa937d03","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T08:40:07.423151Z","strongest_claim":"PaAno achieved state-of-the-art performance, significantly outperforming existing methods, including those based on heavy architectures, on both univariate and multivariate time-series anomaly detection across various range-wise and point-wise performance measures.","one_line_summary":"PaAno uses patch-based 1D CNN embeddings trained with triplet and pretext losses to achieve state-of-the-art time-series anomaly detection on the TSB-AD benchmark for both univariate and multivariate data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the learned patch embeddings reliably separate normal from anomalous temporal patterns and that comparison to training-set normal patches produces a valid anomaly score without additional calibration or post-hoc tuning.","pith_extraction_headline":"A patch-based CNN method for time-series anomaly detection surpasses complex models on benchmarks."},"references":{"count":37,"sample":[{"doi":"10.1145/3394486.3403392","year":null,"title":"doi: 10.1145/3394486.3403392. Shaojie Bai, J. Zico Kolter, and Vladlen Koltun. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling.arXiv preprint arXiv:1803.0","work_id":"558a53f5-6cb5-46ac-8be8-07cb57895f23","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1109/cvpr.2019.00982","year":2019,"title":"Mvtec AD - A comprehensive real-world dataset for unsupervised anomaly detection","work_id":"2f0b7c01-862e-43bb-8146-dd8cf0e3993a","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Paul Boniol, John Paparrizos, Themis Palpanas, and Michael J","work_id":"fd8c9339-7c3b-42bf-91f0-fce27c5eb5da","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.14778/3467861.3467865","year":null,"title":"Paul Boniol, Qinghua Liu, Mingyi Huang, Themis Palpanas, and John Paparrizos","work_id":"0d98a3fa-d47a-4e15-93e0-062eebc68f37","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Breunig, Hans-Peter Kriegel, Raymond T","work_id":"ff9c9c04-2bea-4ced-a04e-19d0004a386a","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":37,"snapshot_sha256":"6387d952cd8bc76fd0a10784b10f9844e8aff57402583e9db9e629f02bdf0ea3","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"e29e1e4daf5e6535de72ec0d7ceaa7417606b2b28c4e1d1e88c3d84c985ef70f"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}