{"paper":{"title":"The Sample Complexity of Multiple Change Point Identification under Bandit Feedback","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Sample complexity for locating multiple change points under bandit feedback is governed jointly by jump sizes and their relative positions.","cross_cats":["cs.LG","math.ST","stat.TH"],"primary_cat":"stat.ML","authors_text":"Maximilian Graf, Victor Thuot","submitted_at":"2026-05-13T09:35:19Z","abstract_excerpt":"We study multiple change point localization under bandit feedback. An unknown piecewise-constant function on a compact interval can be queried sequentially at adaptively chosen inputs, and each query returns a noisy evaluation of the function. The goal is to identify a prescribed number of discontinuities, known as change points, within a target precision $\\eta$ and confidence level $1-\\delta$, while using as few samples as possible. We propose an adaptive algorithm that first detects intervals likely to contain change points and then refines their locations to precision $\\eta$. We establish n"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We demonstrate, both empirically and theoretically, that for general δ and η, the complexity is jointly governed by the jumps and the relative positions of the change points.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The number of change points is known in advance and the underlying function is exactly piecewise constant with jumps of unknown but fixed magnitudes; if the number is unknown or the function has smooth transitions instead of sharp jumps, the detection and refinement phases may fail to achieve the stated precision.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"An adaptive bandit algorithm for multiple change-point localization achieves non-asymptotic sample bounds jointly controlled by jump magnitudes and change-point spacing for any fixed δ and η.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Sample complexity for locating multiple change points under bandit feedback is governed jointly by jump sizes and their relative positions.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f040c4a91066f728ce927b0c256fda7fd2c539b7b941a07498ca0bc24fb60ebe"},"source":{"id":"2605.13252","kind":"arxiv","version":1},"verdict":{"id":"8a5a5687-0378-405d-9870-55fa974f2aae","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T17:44:18.218992Z","strongest_claim":"We demonstrate, both empirically and theoretically, that for general δ and η, the complexity is jointly governed by the jumps and the relative positions of the change points.","one_line_summary":"An adaptive bandit algorithm for multiple change-point localization achieves non-asymptotic sample bounds jointly controlled by jump magnitudes and change-point spacing for any fixed δ and η.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The number of change points is known in advance and the underlying function is exactly piecewise constant with jumps of unknown but fixed magnitudes; if the number is unknown or the function has smooth transitions instead of sharp jumps, the detection and refinement phases may fail to achieve the stated precision.","pith_extraction_headline":"Sample complexity for locating multiple change points under bandit feedback is governed jointly by jump sizes and their relative positions."},"references":{"count":45,"sample":[{"doi":"","year":2014,"title":"Adaptive sensing performance lower bounds for sparse signal detection and support estimation , author=. Bernoulli , volume=. 2014 , publisher=","work_id":"0a407f93-91c5-434a-b1a5-876f7cc7ad41","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"The 28th International Conference on Artificial Intelligence and Statistics , year=","work_id":"8ac2bab9-4e49-430e-8d6d-36cd08f5b276","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Proceedings of the 42nd International Conference on Machine Learning , pages =","work_id":"c19854bc-64d3-4c14-a768-a6933ed9794f","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2016,"title":"Conference on Learning Theory , pages=","work_id":"1bb1909a-9054-4311-873c-f4a3a3fc0873","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Safety Aware Changepoint Detection for Piecewise i.i.d. Bandits , author=. The 38th Conference on Uncertainty in Artificial Intelligence , year=","work_id":"39ec369c-879c-48ff-ad98-ad8dd170a640","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":45,"snapshot_sha256":"80b8617f22001bedb7082ed737331d5c71acb93de8761974fc7526cebd756981","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"ef9575b1c42969d1947202da7914291feb1a3a56af72c3a78b7eb6455e1d75f5"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}