Neural network verification researcher at Vanderbilt: Star-set reachability analysis for GNNs and new architectures
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-02) 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—
Not yet scored: this profile's data covers only 2 reviewed arXiv preprints, with no total publication count or venue history available. This axis would measure cumulative research output, such as papers per year across the full career.
Impact—
Not yet scored: citation counts, h-index, and venue-level influence are absent from this snapshot. This axis would measure how widely the researcher's verification methods are taken up by other groups through citations and downstream use.
Focus0/100
The focus score is 0, computed from a field distribution that splits exactly 50/50 between 'formal verification of machine learning (computer science / cyber-physical systems)' and 'machine learning formal verification (GNN verification for power systems)'. Both halves attack the same underlying problem of verifying neural networks formally and share methods such as reachability analysis and Star sets, so the work reads as tightly clustered even though the even two-field split drives the low score.
Collaboration55/100
The collaboration score is 55, based on 15 distinct co-authors in the current bibliometric snapshot. The recurring collaborator list includes Taylor T. Johnson, alongside Anne M. Tumlin, Ben Wooding, Samuel Sasaki, Muhammad Usama Zubair, Navid Hashemi, Hongchao Zhang, Waseem Abbas, Ipek Oguz, Meiyi Ma, Zhenxuan Shao, and Tyler Derr. On both cataloged 2026 papers the researcher sits in a middle author position without corresponding status, pointing to team-based rather than solo work.
Activity—
Not yet scored: yearly publication counts cannot be computed from 2 same-year (2026) preprints. This axis would measure research tempo, such as outputs per year and continuity of activity in recent years.
Reproducibility100/100
Reproducibility scores 100, the mean across the 2 papers with reproducibility reviews, both rated 'high' readiness. The GraphStar-set reachability method is benchmarked against CORA, and NNV3 consolidates weight-perturbation, video, GNN, probabilistic, and fairness verification into a single MATLAB tool, both supporting independent reuse. Reviews flag no reproducibility barriers — noted issues concern PF feature design and selection, and small evaluation scale, not artifact availability.
Research summary
Diego Manzanas Lopez is a researcher at Vanderbilt University working on formal verification of machine learning. The platform's field distribution splits the portfolio exactly 50/50 between formal verification of machine learning in cyber-physical systems and GNN verification for power systems. Two 2026 arXiv preprints anchor the catalog record. The first, on reachability-based formal verification of graph neural networks with node and edge features, claims the first edge-aware verification of GNNs via GraphStar sets; Evolve's review finds the method sound with strong CORA comparisons while flagging PF feature design and selection issues. The second, NNV3, unifies weight-perturbation, video, GNN, probabilistic, and fairness verification in a single MATLAB tool; the integration is judged credible but evaluations stay small. Both papers earn a 'high' reproducibility-readiness rating, giving a reproducibility mean of 100. The researcher works with 15 distinct co-authors, including Taylor T. Johnson, Waseem Abbas, and Tyler Derr, in middle-author, non-corresponding positions on both papers. Output, impact, and activity metrics are not yet available for this profile.
Collaboration pattern
With 15 distinct co-authors (collaboration score 55), Manzanas Lopez publishes in a team setting. Frequent collaborators include Taylor T. Johnson, Anne M. Tumlin, Ben Wooding, Samuel Sasaki, Muhammad Usama Zubair, Navid Hashemi, Hongchao Zhang, Waseem Abbas, Ipek Oguz, Meiyi Ma, Zhenxuan Shao, and Tyler Derr. On both cataloged papers the researcher holds a middle author position and is not the corresponding author, consistent with contributing within a larger Vanderbilt verification group.
Research keywords
Highlights
First edge-aware formal verification of GNNs via GraphStar sets
NNV3: one MATLAB tool spanning five verification frontiers
Timeline
- 2026Co-authored NNV3, a MATLAB tool unifying weight-perturbation, video, GNN, probabilistic, and fairness verification under one framework.
- 2026Co-authored reachability-based verification of graph neural networks with node and edge features, claiming the first edge-aware method via GraphStar sets.
- 2026Both preprints entered the Evolve catalog with 'high' reproducibility readiness, yielding a reproducibility mean of 100 across 2 reviewed papers.
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.
- Dimension scores are derived from the platform's OpenAlex-based bibliometric snapshot only; citation counts, h-index, and venue metrics are not available.
- No ORCID is linked; the profile is name-resolved from public bibliographic data, so same-name researchers could in principle be conflated.
- The record covers only 2 reviewed papers (both 2026 arXiv preprints), which is why output, impact, and activity are marked as missing.
- Country and affiliation history are not provided; institutional scope is limited to the single listed Vanderbilt University affiliation.