Irina Tezaur

—

Unclaimed · AI-structured profileOpenAlex

Machine-learned reduced models coupled via overlapping Schwarz domain decomposition in computational mechanics

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-09-20) 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—

No publication-volume data is available in this snapshot. The output axis would measure the number, venues, and types of publications over time; with only two reviewed 2026 preprints, no output score can be computed.

Impact—

Citation counts, h-index, and other influence indicators are absent from the provided data. The impact axis would measure how widely this work is cited and taken up by others; no such metrics can be reported yet.

Focus0/100

The focus score is 0, and the raw field distribution explains it: the two cataloged papers split evenly, 0.5 and 0.5, between scientific machine learning / computational mechanics and computational science and engineering (numerical PDEs). The two labels are adjacent and both papers share the overlapping Schwarz coupling backbone, so on this very small two-paper sample the zero reflects an even split across field tags rather than clearly scattered research.

Collaboration36/100

The collaboration score is 36, based on 5 distinct coauthors across the reviewed papers. The recurring network is compact — Trishit Mondal, Anthony Gruber, George Chumbipuma, Alejandro Diaz, and Beatrice Riviere — and Tezaur sits in the middle author position on both 2026 preprints without corresponding authorship, indicating a contributing role in a small, stable team rather than leadership of a large distributed collaboration.

Activity—

Recent publication cadence cannot be assessed from two same-year preprints. The activity axis would measure ongoing research output and engagement over recent years; the underlying data is not available in the current snapshot.

Reproducibility55/100

Reproducibility scores a mean of 55 across the two reviewed papers, both tagged with medium reproducibility readiness. The RL-switching paper reports strong benchmark results but from a single case with unquantified training costs, while the NINN coupling paper documents precise error levels (about 0.05-0.1% at Peclet number 10^6) yet shows no demonstrated computational benefit and releases no code — reproducible in principle, not yet in practice.

Research summary

Irina Tezaur is a researcher in scientific machine learning and computational mechanics whose Evolve catalog centers on coupling machine-learned reduced models with classical domain decomposition. Both cataloged papers appeared as arXiv preprints in 2026 and share one methodological skeleton: the overlapping Schwarz alternating method. In one line of work, operator-inference reduced order models are combined with reinforcement learning — a deep Q-network — to switch online between full-order and reduced-order solvers inside a Schwarz-coupled system; the reviewed study reports strong benchmark gains, though the evidence rests on a single case with unquantified training costs. The second paper couples pre-trained, frozen-weight numerics-informed neural networks (NINNs) to full-order subdomain solvers, reaching roughly 0.05-0.1% error on an advection-diffusion problem at Peclet number 10^6; Evolve characterizes it as a careful feasibility study whose computational benefit is undemonstrated and whose code is not released. The cataloged field distribution splits evenly between the two field labels, and the coauthor network is compact: five frequent collaborators, with Tezaur as a middle author on both papers. Output, impact, and activity metrics are not yet available in this snapshot, so this profile should be read as an early-stage, name-resolved bibliographic portrait built from the two reviewed papers.

Collaboration pattern

The cataloged footprint is a compact five-person coauthor network — Trishit Mondal, Anthony Gruber, George Chumbipuma, Alejandro Diaz, and Beatrice Riviere — underlying a collaboration dimension score of 36. On both 2026 preprints Tezaur appears as a middle author rather than first or corresponding, consistent with a contributing role inside a small, stable team spanning scientific machine learning and computational mechanics.

Research keywords

domain decompositionSchwarz alternating methodhybrid couplingreduced order modelingoperator inferencereinforcement learningdeep Q-networknumerics-informed neural networksphysics-informed neural networksadvection-diffusion equationhigh Peclet numbercomputational mechanics

Highlights

Reinforcement-learning switching between full-order and reduced-order solvers
On a 2026 arXiv preprint, Tezaur co-developed an adaptive hybrid coupling scheme that uses a deep Q-network to switch online between full-order and operator-inference reduced-order solvers within the overlapping Schwarz alternating method; the reviewed study reports strong benchmark gains, tempered by single-case reporting and unquantified training costs.
Frozen-weight numerics-informed neural networks coupled at Peclet number 10^6
A second 2026 preprint couples pre-trained, frozen-weight numerics-informed neural networks (NINNs) to full-order subdomain solvers through the overlapping Schwarz alternating method, reaching roughly 0.05-0.1% error on an advection-diffusion problem at Peclet number 10^6; Evolve rates it a careful feasibility study with undemonstrated computational benefit and no released code.
A consistent thread: Schwarz-based hybrid coupling
Both cataloged papers — the RL-switched operator-inference coupling and the NINN-based coupling — build on the same overlapping Schwarz alternating backbone, pointing to a focused methodological thread that pairs machine-learned components with classical domain decomposition, with medium reproducibility readiness recorded for each paper.

Timeline

2026Co-authored 'Adaptive hybrid coupling with operator inference, the overlapping Schwarz alternating method and reinforcement learning', an arXiv preprint with a middle author position.
2026Co-authored 'Hybrid coupling with numerics-informed neural networks and the overlapping Schwarz alternating method', an arXiv preprint, again as a middle author.
2026Both preprints entered the Evolve review catalog with medium reproducibility readiness, spanning two adjacent field labels split evenly in the recorded field distribution.

Papers in the Evolve catalog

Frequent co-authors

Trishit Mondal×1 Anthony Gruber×1 George Chumbipuma×1 Alejandro Diaz×1 Beatrice Riviere×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.

Bibliometric dimensions are computed from a small snapshot: only two 2026 arXiv preprints are reviewed, so the output, impact, and activity axes lack sufficient data.
No institution, country, or ORCID was provided; the profile is name-resolved from public bibliographic data and may be affected by name ambiguity.
The reproducibility score (55) is a mean over two reviewed papers, both with medium reproducibility readiness and no released code noted for the NINN paper.
Field distribution and coauthor counts reflect only the Evolve-reviewed subset, not the researcher's complete publication record.