Evolve.Pub
Sign inScan your paper
RC

Romain Claret

University of Neuchâtel · Switzerland

Unclaimed · AI-structured profileORCID OpenAlex

Neuroevolution researcher at Neuchâtel: first-authored ES-HyperNEAT work on indirect encoding and hyperparameter tuning.

4

papers here

3

reviewed by Evolve

4

first-author

2

corresponding

—

works · OpenAlex

1

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-17) 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: insufficient dataOutputImpact: insufficient dataImpactFocus: 0/100FocusCollaboration: 41/100CollaborationActivity: 70/100ActivityReproducibility: 70/100Reproducibility

Output—

Not scored: the data pipeline returned no reliable output metrics for this profile yet. This axis would measure total publication count, output trajectory across years, and spread across venues and output types.

Impact—

Not scored: citation-based measures are too sparse at this stage (h-index 1, i10-index 0) to support an estimate. This axis would measure total citations, citation velocity, and uptake of the work by follow-up research.

Focus0/100

Focus scores 0 on the radar, and the raw field distribution shows why: work splits evenly into thirds — evolutionary machine learning, machine-learning hyperparameter optimization, and evolutionary computation, all under the neuroevolution umbrella. The spread is across adjacent subfields rather than scattered topics, but no single concentration dominates the record yet.

Collaboration41/100

Collaboration scores 41. The recent window records 7 distinct co-authors, with recurring partners including Paul Cotofrei, Kilian Stoffel, Michael O'Neill, Arthur Gygax, Pascal Felber, and Michael Palma Mendes. Catalog papers are consistently first-authored, two of them also corresponding, pointing to a small stable network and a driving personal role rather than large distributed teams.

Activity70/100

Activity scores 70, the strongest axis in the record. countsByYear shows 1 indexed work in 2024 jumping to 9 in 2026, with 10 works in the last three years. Four of the 2026 outputs are first-authored arXiv preprints already reviewed in the Evolve catalog, confirming concentrated, ongoing output rather than a one-off historical spike.

Reproducibility70/100

Reproducibility scores 70, averaged over 3 scored papers (reproMean 70). Evolve rates the spatially-partitioned-experts study high for its controlled, coverage-based design; the palette-evolution, early-stopping, and TPE studies are rated medium, with flagged issues such as non-budget-matched baselines, single-task validation, and a missing test-set comparison.

Research summary

Romain Claret is an early-career researcher at the University of Neuchâtel (Switzerland) working at the intersection of neuroevolution, evolutionary computation, and machine-learning hyperparameter optimization. His OpenAlex record is compact — h-index 1, i10-index 0, and 10 works in the last three years — but 2026 marks a concentrated output phase, with 9 indexed works and four first-authored arXiv preprints reviewed in the Evolve catalog, three of them as corresponding author. The work systematically dissects ES-HyperNEAT and indirectly encoded substrates: spatially partitioned experts that lift MNIST accuracy from 21% to 43% by countering central bias, a Tree-structured Parzen Estimator tuning study reaching 29.0% best MNIST accuracy, data-driven early-stopping thresholds derived from fitness dynamics, and bio-inspired palette strategies for activation-function discovery in CPPN substrates. Topic profiling places him chiefly in Evolutionary Algorithms and Applications (10), Neural Networks and Applications (8), and Metaheuristic Optimization Algorithms Research (7). Evolve's reproducibility assessment averages 70 across three scored papers, with the spatial-partitioning study rated high. The record reflects an emerging, method-focused agenda rather than an established citation footprint.

Collaboration pattern

Collaboration is compact and local: 7 distinct co-authors in the recent window, scoring 41 on the radar. Recurring partners include Paul Cotofrei, Kilian Stoffel, Michael O'Neill, Arthur Gygax, Pascal Felber, and Michael Palma Mendes. All four catalog papers are first-authored and two are also corresponding, consistent with a driving individual role within a small stable group at Neuchâtel rather than large multi-institution teams.

Research keywords

neuroevolutionES-HyperNEATHyperNEAT and CPPNindirect encodingactivation function discoveryhyperparameter optimizationBayesian optimization (TPE)early stoppingfitness trajectory analysisMixture-of-Expertsspatial partitioningmeta-learning

Highlights

Timeline

  1. 2024First indexed work appears in the OpenAlex record (1 output in 2024), marking the start of citable research activity.
  2. 2026Output accelerates sharply: 9 works indexed in a single year, the highest point of the record's countsByYear series.
  3. 2026"Breaking the Central Bias" introduces spatially partitioned experts, lifting ES-HyperNEAT MNIST accuracy from 21% to 43%; Evolve rates its reproducibility high.
  4. 2026The TPE hyperparameter study (29.0% best MNIST accuracy) and data-driven early-stopping thresholds round out a four-paper first-authored arXiv run.

Papers in the Evolve catalog

Frequent co-authors

Paul Cotofrei×4Michael O'Neill×2Kilian Stoffel×2Arthur Gygax×2Michael O'Neill×2Pascal Felber×2Michael Palma Mendes×1

Is this you? Claim this profile

Enter the email you published with. We'll send a one-time claim link; after claiming you can edit the introduction or hide this page. Requests from a mismatched address go to manual review.

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, counts by year, topics) come from OpenAlex only; coverage of venues and citations may be incomplete.
  • Output and impact radar axes are unscored: citation data are too sparse (h-index 1, i10-index 0) to support a meaningful estimate.
  • The reproducibility score averages Evolve assessments of 3 papers; it does not cover the full output record.
  • Catalog coverage is limited to 4 arXiv preprints reviewed by Evolve; the profile is name-resolved from public bibliographic data via ORCID 0000-0002-5612-8471.
Romain Claret — Researcher Profile | Evolve.Pub