Journal article · 2026 · Open access

HC-GPUCB: Hybrid Concentration-Aware Gaussian Process Upper Confidence Bound for Bayesian Optimization with Tight Regret Guarantees

Baobing Zhang, Wanxin Sui

IEEE Transactions on Artificial Intelligence, 1–13 · Published 1 January 2026 · DOI 10.1109/tai.2026.3734017

Open full textCite this paper

In brief

“HC-GPUCB: Hybrid Concentration-Aware Gaussian Process Upper Confidence Bound for Bayesian Optimization with Tight Regret Guarantees” is a 2026 journal article by Baobing Zhang and Wanxin Sui, published in IEEE Transactions on Artificial Intelligence. OpenAlex counts 0 citations as of 19 November 2026. It is open access under a CC BY licence: the full text is free to read at IEEE Transactions on Artificial Intelligence.

Citations
0
Authors
2

Abstract

Bayesian optimization with the Gaussian process Upper Confidence Bound (GP-UCB) enjoys strong regret guarantees but scales poorly: over a budget ofTevaluations, the cubic Cholesky factorization at roundtinflates the total cost toO(T4). A recent analysis by Iwazaki (NeurIPS 2025) sharpens GP-UCB’s Bayesian cumulative regret to order$\sqrt{T}$, up to logarithmic factors, by exploiting that its queries automatically concentrate around the global maximizer. We promote this concentration from a passive analytical tool to an active design principle: HC-GPUCB runs vanilla GP-UCB globally for a warm-up of order$\sqrt{T}$iterations, then restricts the search domain and active dataset to a provably correct local ball around the post-warm-up incumbent. Correctness of that ball follows from the sub-optimality-to-neighbourhood lemma of Iwazaki (their Lemma 20) once it is re-centred on the algorithmically available incumbent; on that basis we prove regret preservation at that same rate, together with a structural-complexity result exposing the localized GP-fit substrate to engineering speedups. We complement HC-GPUCB with Checkpoint GP-UCB— a drop-in variant using logarithmic-checkpoint hyperparameter refits with incremental posterior conditioning—and integrate the two as HC-Checkpoint-GPUCB. Empirically, on four standard benchmarks (d ∊ {2, 3, 5, 6},T≤ 3000), Checkpoint GP-UCB delivers 3.5–7.1× wall-clock speedups atT=500 and grows to 16.3× atT=1000 on a controlled single-function experiment; HC-GPUCB matches vanilla regret on the harder benchmarks (Wilcoxonp≥0.77 on Hartmann-6 and Ackley-5); and HC-Checkpoint-GPUCB is the only entry that scales toT=3000 on every benchmark, its fixed-known-kernel specialization carrying that guarantee. Code:https://github.com/baobingzhang/HC-GPUCB.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Zhang, B., & Sui, W. (2026). HC-GPUCB: Hybrid Concentration-Aware Gaussian Process Upper Confidence Bound for Bayesian Optimization with Tight Regret Guarantees. IEEE Transactions on Artificial Intelligence, 1–13. https://doi.org/10.1109/tai.2026.3734017

MLA 9

Zhang, Baobing, and Wanxin Sui. “HC-GPUCB: Hybrid Concentration-Aware Gaussian Process Upper Confidence Bound for Bayesian Optimization with Tight Regret Guarantees.” IEEE Transactions on Artificial Intelligence, 2026, pp. 1–13. https://doi.org/10.1109/tai.2026.3734017.

Chicago

Zhang, Baobing and Wanxin Sui. 2026. “HC-GPUCB: Hybrid Concentration-Aware Gaussian Process Upper Confidence Bound for Bayesian Optimization with Tight Regret Guarantees.” IEEE Transactions on Artificial Intelligence: 1–13. https://doi.org/10.1109/tai.2026.3734017.

Harvard

Zhang, B. and Sui, W. (2026) ‘HC-GPUCB: Hybrid Concentration-Aware Gaussian Process Upper Confidence Bound for Bayesian Optimization with Tight Regret Guarantees’, IEEE Transactions on Artificial Intelligence, pp. 1–13. doi:10.1109/tai.2026.3734017.

IEEE

B. Zhang and W. Sui, “HC-GPUCB: Hybrid Concentration-Aware Gaussian Process Upper Confidence Bound for Bayesian Optimization with Tight Regret Guarantees,” IEEE Transactions on Artificial Intelligence, pp. 1–13, 2026, doi: 10.1109/tai.2026.3734017.

BibTeX

@article{zhang2026gpucb,
  title     = {{HC-GPUCB: Hybrid Concentration-Aware Gaussian Process Upper Confidence Bound for Bayesian Optimization with Tight Regret Guarantees}},
  author    = {Zhang, Baobing and Sui, Wanxin},
  journal   = {IEEE Transactions on Artificial Intelligence},
  year      = {2026},
  pages     = {1--13},
  doi       = {10.1109/tai.2026.3734017},
  url       = {https://doi.org/10.1109/tai.2026.3734017}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Social Sciences › Decision Sciences › Management Science and Operations Research › Advanced Bandit Algorithms Research

  • Bayesian optimization
  • regret bounds
  • computational complexity reduction
  • checkpointing

Work with this paper

Most-cited papers in Decision Sciences

All highly cited Decision Sciences papers

Questions about this paper

How do I cite this paper in APA style?

Zhang, B., & Sui, W. (2026). HC-GPUCB: Hybrid Concentration-Aware Gaussian Process Upper Confidence Bound for Bayesian Optimization with Tight Regret Guarantees. IEEE Transactions on Artificial Intelligence, 1–13. https://doi.org/10.1109/tai.2026.3734017

How many times has this paper been cited?

0 times, according to OpenAlex on 19 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at IEEE Transactions on Artificial Intelligence: sign in and open the full text from this page.

Who wrote this paper?

Baobing Zhang, Wanxin Sui.

What is the DOI of this paper?

10.1109/tai.2026.3734017, permanently at https://doi.org/10.1109/tai.2026.3734017. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 19 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.