{"paper":{"title":"ALMAB-DC: Active Learning, Multi-Armed Bandits, and Distributed Computing for Sequential Experimental Design and Black-Box Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"ALMAB-DC pairs Gaussian process active learning with multi-armed bandit allocation and asynchronous distributed scheduling to cut regret and wall-clock time on expensive black-box tasks.","cross_cats":["stat.CO","stat.ME","stat.ML"],"primary_cat":"cs.LG","authors_text":"Foo Hui-Mean, Yuan-chin I Chang","submitted_at":"2026-03-22T11:47:20Z","abstract_excerpt":"Sequential experimental design under expensive, gradient-free objectives is a central challenge in computational statistics: evaluation budgets are tightly constrained and information must be extracted efficiently from each observation. We propose \\textbf{ALMAB-DC}, a GP-based sequential design framework combining active learning, multi-armed bandits (MAB), and distributed asynchronous computing for expensive black-box experimentation. A Gaussian process surrogate with uncertainty-aware acquisition identifies informative query points; a UCB or Thompson-sampling bandit controller allocates eval"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"ALMAB-DC achieves lower simple regret than Equal Spacing, Random, and D-optimal designs on statistical tasks, 93.4% CIFAR-10 accuracy outperforming BOHB and Optuna, 36.9% drag reduction, 50% RL improvement, and 7.5x speedup at K=16, with all advantages statistically significant.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The Gaussian process surrogate accurately models the black-box objective and that the UCB/Thompson sampling bandit controller with asynchronous scheduler effectively allocates evaluations without significant overhead or synchronization issues.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ALMAB-DC integrates Gaussian process active learning with multi-armed bandit allocation and distributed asynchronous computing to achieve lower regret and faster wall-clock performance in sequential experimental design.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ALMAB-DC pairs Gaussian process active learning with multi-armed bandit allocation and asynchronous distributed scheduling to cut regret and wall-clock time on expensive black-box tasks.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"9505a03ab8dd0d292de9eb3e470795325d8126ea178c754c495e7fe7ffd07437"},"source":{"id":"2603.21180","kind":"arxiv","version":4},"verdict":{"id":"25452ec5-db9c-4cb2-b0a0-73640a864df8","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T07:12:13.237692Z","strongest_claim":"ALMAB-DC achieves lower simple regret than Equal Spacing, Random, and D-optimal designs on statistical tasks, 93.4% CIFAR-10 accuracy outperforming BOHB and Optuna, 36.9% drag reduction, 50% RL improvement, and 7.5x speedup at K=16, with all advantages statistically significant.","one_line_summary":"ALMAB-DC integrates Gaussian process active learning with multi-armed bandit allocation and distributed asynchronous computing to achieve lower regret and faster wall-clock performance in sequential experimental design.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The Gaussian process surrogate accurately models the black-box objective and that the UCB/Thompson sampling bandit controller with asynchronous scheduler effectively allocates evaluations without significant overhead or synchronization issues.","pith_extraction_headline":"ALMAB-DC pairs Gaussian process active learning with multi-armed bandit allocation and asynchronous distributed scheduling to cut regret and wall-clock time on expensive black-box tasks."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.21180/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"b51a747aa0fdc5041e199a619c3e1497d24ee5fa4e7372409adaaeb0b37005f3"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}