Neuroevolution researcher at University of Neuchâtel renewing a two-decade record around ES-HyperNEAT
4
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
7
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—
Total publication counts and venue-level output are not available in the current snapshot, so this axis is unscored. It would measure aggregate output volume, the mix of venues, and output consistency across the full career record.
Impact—
Citation-based impact indicators are not yet available in the provided data, so this axis is unscored. It would measure citation accumulation, the reach of citing works, and uptake of his ES-HyperNEAT methods by other groups.
Focus0/100
The focus axis scores 0/100, reflecting exceptional topical breadth rather than a narrow specialty: the topic profile splits across five clusters — data mining algorithms (14 works), time series analysis and forecasting (11), evolutionary algorithms (10), neural networks (7), and logic and reasoning (7) — while the 2026 catalog work divides evenly, one-third each, across evolutionary machine learning, hyperparameter optimization, and evolutionary computation.
Collaboration41/100
Collaboration scores 41/100. The snapshot records 7 distinct recent coauthors, with recurring partners including Romain Claret, Michael O'Neill, Kilian Stoffel, Arthur Gygax, Pascal Felber, and Michael Palma Mendes. All four Evolve-reviewed 2026 preprints place him in a middle author position and none as corresponding author — a pattern of contributing specialist expertise within small teams rather than leading large consortia.
Activity81/100
Activity scores 81/100 on a clear late-career acceleration. The yearly record shows sporadic output from 1998 to 2015 (peaking at 3 works in 2005 and 2013), a gap from 2016 to 2022, then 3 works in each of 2023, 2024 and 2025 and 9 works in 2026 — the highest single year on record — for 15 works over the last three years.
Reproducibility70/100
Reproducibility scores 70/100, with three papers assessed at a mean of 70. Within the Evolve catalog, 'Breaking the Central Bias' reaches high reproducibility readiness for its well-controlled spatial-partitioning experiments, while the TPE-tuning and early-stopping papers are rated medium, with verdicts noting honest bug validation but budget-mismatched or missing baselines.
Research summary
Paul Cotofrei is a researcher at the University of Neuchâtel (Switzerland) whose recent work centers on neuroevolution — the evolution of neural network architectures and encodings. His OpenAlex record spans more than two decades, with topical clusters in data mining algorithms (14 works), time series analysis and forecasting (11), evolutionary algorithms (10), neural networks (7), and logic and reasoning (7). After a quiet stretch following 2015, output has rebounded sharply: 15 works in the last three years, including 9 indexed in 2026, his most productive year on record. Four 2026 arXiv preprints reviewed by Evolve form a coherent ES-HyperNEAT research program: spatially partitioned experts lifting MNIST accuracy from 21% to 43%, Tree-structured Parzen Estimator tuning reaching 29.0% best accuracy, data-driven early-stopping thresholds from fitness dynamics, and bio-inspired palette evolution in indirectly encoded substrates. Across these he authors in middle positions within small teams (7 distinct coauthors), and three papers carry a mean reproducibility score of 70. His h-index is 7 and i10-index 6 in the OpenAlex snapshot.
Collaboration pattern
The collaboration footprint is compact: 7 distinct recent coauthors and a collaboration score of 41/100. Recurring partners include Romain Claret, Michael O'Neill, Kilian Stoffel, Arthur Gygax, Pascal Felber, and Michael Palma Mendes. All four Evolve-reviewed 2026 preprints are middle-authored, non-corresponding contributions, consistent with a specialist role inside small academic teams at the University of Neuchâtel working with external collaborators.
Research keywords
Highlights
Spatial partitioning lifts ES-HyperNEAT MNIST accuracy from 21% to 43%
TPE tuning pushes ES-HyperNEAT to 29.0% best MNIST accuracy
Early-stopping thresholds for ES-HyperNEAT derived from fitness dynamics
Timescale compatibility shapes bio-inspired palette evolution
Timeline
- 1998First year with a recorded publication in the OpenAlex counts-by-year series (1 work).
- 2005Early-career peak with 3 works recorded, the high point of a steady 2002–2008 publishing run.
- 2013Another 3-work year, matching the pre-2015 productivity peak.
- 2015Last year of the first active phase before a multi-year gap (2 works recorded).
- 2023Return to regular publishing, with 3 works recorded in each of 2023, 2024 and 2025.
- 2026Most productive year on record: 9 works, including four Evolve-reviewed ES-HyperNEAT/neuroevolution preprints.
Papers in the Evolve catalog
Bio-Inspired Palette Evolution in Indirectly Encoded Substrates: Timescale Compatibility Shapes Activation Function Discovery
Early-Stopping Thresholds for ES-HyperNEAT: A Data-Driven Approach from Fitness Dynamics
Breaking the Central Bias: Spatially Partitioned Experts for Coordinate-Based Neuroevolution
Investigating Hyperparameter Optimization and Transferability for ES-HyperNEAT: A TPE Approach
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 7, i10-index 6, counts by year, topics) come from an OpenAlex snapshot and inherit its coverage limits.
- Output and impact dimensions are unscored because aggregate citation and venue data are absent from the provided snapshot.
- Evolve paper-level assessments cover four 2026 arXiv preprints only; all are middle-authored and none is a corresponding-author paper.
- The profile is name-resolved from public bibliographic data (OpenAlex identity; ORCID 0000-0002-4103-5467); disambiguation against same-name authors has not been independently verified.