Brian Bullins

IMDEA Software Institute · United States

Unclaimed · AI-structured profileOpenAlex

Optimization theorist resolving open problems in first-order oracle complexity of convex optimization

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Output-volume scoring is not yet available in this snapshot; this axis would measure aggregate publication counts and productivity over time. Only indirect proxies exist (h-index 9, i10-index 9, 10 indexed works in 2026), and no score is estimated.

Impact—

Citation-impact data is insufficient for scoring this axis, which would measure citation totals, citation velocity, and influence on follow-up work. The snapshot offers no citation figures beyond h-index 9 and i10-index 9, so no impact score is reported.

Focus0/100

This snapshot scores focus at 0/100, a value that sits oddly beside the raw field distribution, which is entirely convex optimization theory (split 0.5/0.5 between two near-identical labels). Topic mass is likewise concentrated: stochastic gradient optimization techniques (39 mentions), advanced optimization algorithms (29), and optimization and variational analysis (15), with minor spillover into multi-agent systems (6) and matrix theory (5). Treat radar scores as indicative and read the raw distribution as the evidence.

Collaboration41/100

Collaboration scores 41/100, consistent with a small, stable co-author pool: 7 distinct co-authors across 16 works in the last three years. Recurring partners David Martínez-Rubio, Cristóbal Guzmán, and Mathieu Molina appear again on both 2026 arXiv papers reviewed by Evolve, where Brian Bullins is a middle-position co-author rather than lead — a pattern of repeat, theory-group collaboration rather than broad network growth.

Activity83/100

Activity is the strongest axis at 83/100. The OpenAlex record spans 2013-2026 with 16 works in the last three years and a clear recent acceleration: 3 indexed works in each of 2024 and 2025, rising to 10 in 2026 — the highest annual count in the snapshot, exceeding earlier peaks of 5 in 2016, 2021, and 2022.

Reproducibility20/100

Reproducibility scores 20/100 (reproMean 20, based on 2 reviewed papers). Both Evolve-reviewed 2026 papers carry a low repro-readiness rating: they are proof-centric theory results — near-optimal acceleration bounds and oracle-complexity characterizations — with no code, artifacts, or empirical benchmarks on record, so independent verification currently rests on the mathematics alone.

Research summary

Brian Bullins, listed at the IMDEA Software Institute, is an optimization theorist working on the first-order oracle complexity of convex optimization. The OpenAlex snapshot (h-index 9, i10-index 9) shows a record concentrated almost entirely in convex optimization theory — the field distribution assigns 100% across two near-identical convex-optimization labels — with topic mass led by stochastic gradient optimization techniques (39 mentions) and advanced optimization algorithms (29). Activity is the strongest radar axis (83/100): 16 works in the last three years and 10 indexed outputs already in 2026, the highest annual count in a record reaching back to 2013. His reviewed 2026 output is theoretically ambitious: two arXiv papers resolve long-standing open problems in nondual convex optimization — near-optimal acceleration for smooth ℓp/ℓq objectives, settling a 2015 open problem of Guzmán, and the first-order oracle complexity of Lipschitz convex optimization, giving Õ(1/T) bounds on the ℓ1-ball via max-regret and sequential fat-shattering dimension, though that algorithm remains exponential-time. Both are middle-author collaborations with recurring co-authors David Martínez-Rubio, Cristóbal Guzmán, and Mathieu Molina, within a small pool of 7 distinct co-authors. Reproducibility is the weakest axis (20/100): proof-centric theory with no code or artifacts on record.

Collaboration pattern

Collaboration is compact and repeat-based: 7 distinct co-authors across 16 works in the last three years, with David Martínez-Rubio, Cristóbal Guzmán, and Mathieu Molina as recurring partners. Both Evolve-reviewed 2026 papers are middle-author team efforts within this circle (Guzmán being the author of the 2015 open problem one of them resolves). The radar collaboration score of 41/100 reflects depth over breadth — dense repeat ties rather than a wide co-author network.

Research keywords

first-order methodsconvex optimizationoracle complexitynondual convex optimizationminimax boundsonline learningstochastic gradient optimizationaccelerated gradient methodsHölder smooth optimizationsequential fat-shattering dimensionvariational analysisdistributed multi-agent optimization

Highlights

Near-optimal acceleration for smooth ℓp/ℓq nondual convex optimization
The 2026 arXiv paper claims near-optimal rates for Hölder-smooth ℓp/ℓq convex first-order oracle optimization, resolving an open problem posed by Guzmán in 2015. The underlying selector machinery is developed in a companion paper, and Evolve rates reproducibility low, as expected for a pure theory result.
First-order oracle complexity of Lipschitz convex optimization in nondual settings
A second 2026 arXiv paper resolves the nonsmooth nondual open question from COLT'15, achieving Õ(1/T) rates on the ℓ1-ball via a max-regret analysis and the sequential fat-shattering dimension, with near-matching bounds — though the algorithm remains exponential-time.

Timeline

2013Earliest indexed output in the OpenAlex snapshot (1 work), the starting point of a record that continues through 2026.
2016First output peak: 5 indexed works, a level later matched only in 2021 and 2022.
2021Sustained mid-period productivity with 5 indexed works, tying 2016 and 2022 as the pre-2026 annual peak.
2024Output steadies at 3 indexed works in each of 2024 and 2025, setting up the 2026 surge.
2026Breakout year: 10 indexed works so far — the highest annual count in the snapshot — including two arXiv papers reviewed by Evolve that resolve long-standing nondual open problems.

Papers in the Evolve catalog

Frequent co-authors

David Martínez-Rubio×2 Cristóbal Guzmán×2 Mathieu Molina×2 David Martínez-Rubio×1 Brian Bullins×1 Cristóbal Guzmán×1 Mathieu Molina×1

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

Bibliometrics come from an OpenAlex snapshot only (identity: openalex; no ORCID linked), so 'Brian Bullins' is name-resolved from public bibliographic data; the listed affiliation is IMDEA Software Institute while the country field reads 'US'.
The output and impact radar axes are marked missing; no publication-volume or citation-impact scores are reported, and no citation totals beyond h-index 9 / i10-index 9 are available.
Radar dims are platform-computed 0-100 scores; the focus score (0) should be read against the raw field distribution, which is 100% convex optimization theory.
Evolve review coverage is limited to two 2026 arXiv papers, both rated low repro-readiness; no code, artifacts, or peer-review outcomes are on record.