Convex optimization theorist resolving long-standing nondual oracle-complexity open problems at Tel Aviv University
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
2
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—
The output axis is not yet scored: OpenAlex coverage for this early-stage profile is incomplete, so reliable publication-volume aggregation is not possible. Once coverage is complete, it would measure total works and per-year output consistency.
Impact—
The impact axis is unscored: the available citation signals (h-index 2, i10-index 1) rest on too few indexed works to be meaningful at this career stage. It would measure citation uptake and the field-wide reach of the research.
Focus0/100
The radar focus score is 0, the lowest available axis. The OpenAlex field distribution sits entirely within convex optimization theory (0.5 and 0.5 across two label variants), yet topic clustering spreads across five clusters — optimization and search problems (12), game theory and voting systems (10), complexity and algorithms in graphs (6), bandit algorithms (6), and ethics and social impacts of AI (4). Topical breadth offsets the concentrated field tags, driving the low score.
Collaboration41/100
Collaboration scores 41/100 on a compact footprint: 7 distinct co-authors over the recent window against 15 works in three years. Both Evolve-reviewed 2026 papers place Molina as last author without corresponding-author designation, and recurring partners David Martínez-Rubio, Brian Bullins, and Cristóbal Guzmán point to small, specialist teams rather than large networks.
Activity81/100
Activity is the strongest axis at 81/100. OpenAlex counts by year climb from 2 works (2020) and 1 (2021) through 4 (2022), 2 (2023), and 4 (2024) to 5 (2025) and a career-high 6 (2026), with 15 works in the last three years — a steep and continuing upswing.
Reproducibility20/100
Reproducibility scores 20, computed from 2 assessed papers with a mean readiness of 20. Both reviewed works rate low: the acceleration result depends on selector machinery in a companion paper, and the nondual Lipschitz algorithm is exponential-time — real limits on practical replication despite the theoretical significance of the claims.
Research summary
Mathieu Molina is a convex optimization theorist at Tel Aviv University whose recent research targets long-standing open questions in the oracle complexity of first-order methods. The Evolve catalog includes two 2026 arXiv papers, both listing Molina as last author: one claims near-optimal acceleration rates for Hölder-smooth ℓp/ℓq nondual convex optimization, resolving Guzmán's 2015 open problem, while the other settles the nonsmooth COLT'15 nondual open question with Õ(1/T) oracle complexity on the ℓ1-ball through max-regret and sequential fat-shattering dimension arguments. OpenAlex coverage is early-stage: an h-index of 2 and i10-index of 1 across 24 works recorded between 2020 and 2026, with 15 works in the last three years and a 2026 peak of 6 — an activity score of 81/100 reflecting steady acceleration. Topic clustering ranks optimization and search problems (12) and game theory and voting systems (10) ahead of graph complexity (6), bandit algorithms (6), and AI ethics (4), while the OpenAlex field distribution sits entirely within convex optimization theory. Collaboration remains compact (41/100; 7 distinct co-authors), anchored by David Martínez-Rubio, Brian Bullins, and Cristóbal Guzmán. Both reviewed papers carry low reproducibility readiness, consistent with a reproducibility score of 20.
Collaboration pattern
Collaboration is compact: 7 distinct co-authors in the recent three-year window and a radar collaboration score of 41/100. Frequent co-authors include David Martínez-Rubio, Brian Bullins, and Cristóbal Guzmán — the latter connected to the 2015 open problem one reviewed paper resolves. Both Evolve-reviewed 2026 papers list Molina as last author without a corresponding-author flag, suggesting a senior role within a small, specialist convex-optimization team rather than large consortium work.
Research keywords
Highlights
Solves the nonsmooth COLT'15 nondual open question
Near-optimal acceleration for Hölder-smooth ℓp/ℓq nondual optimization
Sharply rising research output since 2020
OpenAlex counts by year rise from 1–4 works annually across 2020–2024 to 5 in 2025 and 6 in 2026 — 24 works in total, with 15 in the last three years and an activity score of 81/100.
Timeline
- 2020OpenAlex records 2 indexed works, the start of a profile whose topics now span optimization, game theory, graph complexity, bandit algorithms, and AI ethics.
- 2022Output reaches 4 works — the strongest year to that point, four times the 2021 count of 1.
- 20244 works recorded again, beginning the most productive stretch on record.
- 20255 works recorded, a new annual high at that point.
- 2026Career-high 6 works; two arXiv papers enter the Evolve catalog with Molina as last author — near-optimal ℓp/ℓq nondual acceleration and the Õ(1/T) resolution of the nonsmooth COLT'15 nondual open question.
Papers in the Evolve catalog
Frequent co-authors
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.
- Bibliometric indicators (h-index, i10-index, per-year counts, topics) derive from an OpenAlex identity snapshot and may incompletely cover early-career work.
- Output and impact radar axes are unscored (dimMissing); citation coverage is too thin for reliable scoring at h-index 2.
- Only 2 papers have been Evolve-reviewed, both 2026 arXiv preprints with low reproducibility readiness, so assessments reflect preprint-stage claims.
- Biographical details beyond institution and country are out of scope; this profile is name-resolved from public bibliographic data.