Numerical analyst advancing structure-preserving ML for PDEs and tokamak disruption prediction at Kyushu University
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
10
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) 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 composite is not yet computed for this profile. It would measure overall publication volume and productivity across the full record; at present only coarse indicators exist (h-index 10, i10-index 11), so no output score is stated here.
Impact—
No impact composite is available yet. This axis would capture citation accumulation and scholarly influence across the record; the snapshot currently records only h-index 10 and i10-index 11, without the fuller citation decomposition needed to score it.
Focus0/100
The focus score is 0, reflecting an even 50/50 field split between numerical analysis / scientific machine learning and machine learning for plasma control (fusion disruption prediction). Topic counts confirm the spread: numerical methods for differential equations (44), model reduction and neural networks (28), neural networks and applications (24), computational fluid dynamics and aerodynamics (18), and advanced numerical methods (15).
Collaboration32/100
With a score of 32, collaboration is the narrowest measured axis: only 4 distinct co-authors appear in the recent window. Recurring partners are Teo Deveney, Baige Xu and Takeshi Koshizuka, and on both reviewed 2026 papers Yaguchi is last author without being corresponding author, suggesting focused small-team projects rather than large consortium work.
Activity90/100
Activity scores 90, the strongest axis. The record shows 21 works in the last three years and annual counts of 8 (2024), 7 (2025) and 6 (2026), following a peak of 12 in 2021 and steady presence since 2005 — sustained recent productivity with no long gaps since the early 2010s.
Reproducibility55/100
Reproducibility scores 55, the mean over 2 assessed papers, both rated medium repro-readiness. The Brinkman-penalisation study releases no code despite sound theory, and the disruption-alarm benchmark rests on 3-seed evidence with a calibration gap — reusable in principle, but both would need code release and multi-seed extension to score higher.
Research summary
Takaharu Yaguchi is a researcher at Kyushu University, Japan, whose recorded output splits evenly between two fields: numerical analysis / scientific machine learning and machine learning for plasma control, specifically fusion disruption prediction (a 50/50 split in the field distribution). The snapshot, resolved from OpenAlex and tied to ORCID 0000-0001-9025-6015, records an h-index of 10 and an i10-index of 11, with publications logged from 2005 through 2026. Activity is the strongest measured axis: 21 works in the last three years, including 8 in 2024, 7 in 2025 and 6 in 2026, after a 12-work peak in 2021. Topic modelling concentrates on numerical methods for differential equations (44), model reduction and neural networks (28), neural networks and applications (24), computational fluid dynamics and aerodynamics (18), and advanced numerical methods in computational mathematics (15). Evolve has reviewed two 2026 arXiv papers: a structure-preserving operator-learning study of Brinkman-penalised Hamiltonian PDEs, judged theoretically sound but limited to linear PDEs with no code released, and a horizon-aware tokamak disruption alarm benchmark on DIII-D/EAST, judged plausible but constrained by 3-seed evidence and a calibration gap. Yaguchi is last author on both, in a small-team collaboration pattern with only 4 distinct recent co-authors.
Collaboration pattern
Collaboration is narrow but stable: 4 distinct co-authors in the recent window and a collaboration score of 32. Recurring partners are Teo Deveney, Baige Xu and Takeshi Koshizuka. On both 2026 catalog papers Yaguchi is the last author but not the corresponding author, consistent with a senior collaborator contributing methodological depth to small numerical-analysis and plasma-ML teams rather than leading a large group.
Research keywords
Highlights
Conformal symplectic structure preserved under Brinkman penalisation
Horizon-aware benchmark for tokamak disruption alarms
Timeline
- 2005Earliest works in the OpenAlex record appear, with 2 publications logged in 2005.
- 2010A steady early phase begins: 4 works in 2010, 3 in 2011 and 4 in 2012.
- 2021The most productive year on record, with 12 works published.
- 2024Renewed acceleration: 8 works in 2024, followed by 7 in 2025 and 6 in 2026, with 21 works across the last three years.
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 are derived from OpenAlex only (identity: openalex); venue and citation coverage may be incomplete.
- Output and impact composite scores are not yet computed; only the h-index (10) and i10-index (11) are available.
- The frequent-coauthor list includes the researcher's own name, likely a name-resolution artifact.
- This profile is name-resolved from public bibliographic data (ORCID 0000-0001-9025-6015); only the two reviewed 2026 arXiv papers are individually assessed, and the kanji spelling of the name is unverified.