Researcher at the OR–ML interface, advancing exact and scalable optimal classification trees (interpretable ML)
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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—
Output is unscored: the underlying data — total publication count, venue mix, and yearly production beyond the two reviewed arXiv preprints — is not yet available, so this axis cannot be estimated from current evidence.
Impact—
Impact is unscored: citation counts, h-index, and downstream uptake of the two 2026 preprints are not in the available data, so this axis — which would measure scholarly influence and diffusion — is left blank rather than estimated.
Focus0/100
The radar assigns focus a score of 0, and the field distribution splits evenly (0.5/0.5) between 'machine learning (interpretable models / operations research)' and 'operations research / machine learning (optimal decision trees)'. Both labels, however, describe the same agenda: every reviewed paper targets optimal classification trees, first via an exact junction-tree LP and then via scaling techniques on STreeD — so the formal score reads low while the cataloged work itself is thematically narrow.
Collaboration22/100
Collaboration scores 22/100, reflecting a small footprint: only 2 distinct coauthors appear across the reviewed catalog, and Wenqi Fan is the single recurring collaborator (recorded under the name variants 'Wenqi Fan' and 'WenqiFan'). Both papers are first-authored without a corresponding-author marker, suggesting a lead-contributor role within a compact team rather than a wide collaboration network; the score may understate reach if unreviewed work involves more partners.
Activity—
Activity is unscored: yearly publication counts, submission cadence, and recent-venue data beyond the two 2026 preprints are absent, so this axis — which would measure research tempo and continuity over time — cannot yet be computed.
Reproducibility100/100
Reproducibility scores 100/100, computed over 2 reviewed papers (reproMean 100). Both 2026 arXiv preprints are rated 'high' reproducibility readiness by Evolve's review: the junction-tree paper reports an exact, integral LP formulation with two exact solvers, and the scaling paper layers lossless record deduplication on STreeD with cleanly scoped guarantees and solid ablations — the kind of precise, checkable claims that support independent replication.
Research summary
Jiancheng Tu is a researcher at Hong Kong Polytechnic University working at the interface of operations research and machine learning, focused on optimal classification trees (OCTs) — provably optimal, interpretable decision-tree models. The Evolve catalog lists two 2026 arXiv preprints, both with Tu as first author. The first, 'An Exact Junction-Tree Extended Formulation for Optimal Classification Trees', presents an exact, integral junction-tree linear-programming formulation with reduction techniques and two exact OCT solvers; Evolve's review judges the theory sound and the experiments strong, while noting that depth-5 limits remain. The second, 'Scaling Optimal Classification Trees via Adaptive Feature and Sample Reduction', builds on the STreeD dynamic-programming framework, adding lossless record deduplication and heuristic feature refinement; guarantees are cleanly scoped and ablations are solid, though baselines stay narrow. Bibliometrically, the profile shows an even 0.5/0.5 split between two adjacent field labels, a small collaboration footprint of 2 distinct coauthors — with Wenqi Fan the recurring partner — and a perfect reproducibility reading (reproMean 100 across 2 reviewed papers, each rated high). Output, impact and activity axes are not yet scored due to insufficient data.
Collaboration pattern
Collaboration is compact: 2 distinct coauthors across the reviewed catalog (score 22/100), with Wenqi Fan — recorded under the variants 'Wenqi Fan' and 'WenqiFan' — as the recurring partner. Both papers are first-authored without a corresponding-author marker, and both sit in the optimal-classification-tree space, consistent with a small, focused team at Hong Kong Polytechnic University. The narrow footprint reflects catalog scope; broader networks, if any, are not visible in current data.
Research keywords
Highlights
Exact junction-tree extended formulation for optimal classification trees
Scaling optimal classification trees via adaptive feature and sample reduction
High reproducibility across the reviewed catalog
Both reviewed papers carry a 'high' reproducibility readiness rating, yielding a reproducibility mean of 100 over 2 papers — driven by exact formulations, cleanly scoped guarantees, and reported ablations.
Timeline
- 2026Releases the arXiv preprint 'An Exact Junction-Tree Extended Formulation for Optimal Classification Trees' as first author.
- 2026Releases the arXiv preprint 'Scaling Optimal Classification Trees via Adaptive Feature and Sample Reduction' as first author.
- 2026Both preprints enter the Evolve review catalog, each rated high reproducibility readiness (reproMean 100 over 2 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.
- Bibliometric dimensions are 0-100 scores computed from real bibliometric data; output, impact and activity axes have insufficient data and are not scored.
- Evidence is limited to the two 2026 arXiv preprints reviewed by Evolve; citation counts, venue history and h-index are not available.
- Collaboration metrics cover only the reviewed catalog; the broader co-author network is unknown.
- Profile resolved by name and ORCID (0000-0002-2121-5524) from public bibliographic data; country is not provided in the source record.