Shunnosuke Ikeda

University of Tsukuba · Japan

Unclaimed · AI-structured profileORCID OpenAlex

OR/ML researcher at Tsukuba pairing decision-focused portfolio learning with optimal stratified experiment design

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-15) 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 scored: OpenAlex currently lacks reliable aggregate publication totals for this profile. This axis would measure cumulative research output — total indexed works, venue types, and their distribution over time — none of which can be responsibly estimated here.

Impact—

Not scored: citation evidence is too thin — the profile reports only h-index 4 and i10-index 1. This axis would measure citation-based influence, including uptake, citation trajectory and referencing fields, and remains honestly unestimated here.

Focus0/100

Scores 0, reflecting an even 50/50 split in fieldDist between machine learning and operations research for quantitative finance and optimization-based machine learning for experimental design and survey sampling. Topic counts spread similarly across five areas — risk and portfolio optimization (10), recommender systems (8), supply chain and inventory management (7), consumer market behavior and pricing (7), and optimal power flow distribution (4) — deliberate breadth rather than missing data.

Collaboration36/100

Scores 36, consistent with a compact network: OpenAlex records 5 distinct co-authors over the last three years, with Yuichi Takano, Kensei Nosaka and Tomoka Takei the most frequent partners. Both cataloged 2026 papers place him in a middle authorship position without the corresponding-author role, indicating a contributing role within a small, stable team rather than a corresponding-author lead.

Activity91/100

Scores 91, the profile's strongest axis. OpenAlex counts 22 indexed works in the last three years, with yearly counts climbing from 2 (2023) to 9 (2024), dipping to 2 (2025), then reaching a record 11 in 2026. Even allowing for indexing lag, this pace marks a clear acceleration in research activity.

Reproducibility78/100

Scores 78, based on Evolve's repro-readiness review of 2 cataloged papers (reproMean 78). The optimal multi-way stratification tree paper rates high readiness, while the KKT-based decision-focused learning portfolio paper rates medium — its ETF backtest lacks transaction costs, error bars and significance tests, a gap the Evolve verdict flags explicitly.

Research summary

Shunnosuke Ikeda is an early-career researcher at the University of Tsukuba, Japan, working at the intersection of operations research, machine learning and quantitative decision-making. OpenAlex bibliometrics show a small cumulative record — h-index 4 with i10-index 1 — but a pronounced recent acceleration: 22 indexed works in the last three years, with yearly counts rising from 2 (2023) to 9 (2024) and a record 11 (2026). Two strands define the profile. In machine learning and operations research for quantitative finance, a 2026 arXiv paper proposes KKT-based decision-focused learning for constrained mean-variance portfolio optimization, reporting the best ETF backtest on SR/FW/CDL metrics, though Evolve flags missing transaction costs, error bars and significance tests. In optimization-based machine learning for experimental design and survey sampling, a second 2026 paper builds optimal multi-way decision trees for stratified sampling in online controlled experiments, achieving A/B variance reduction via exact binary optimization, validated so far on two offline datasets. Topic counts extend further into recommender systems (8), supply chain and inventory management (7), consumer market behavior and pricing (7), and optimal power flow distribution (4), indicating deliberate breadth across applied optimization problems. In both cataloged papers Ikeda is a middle author within a compact co-author network of 5 distinct collaborators, anchored by Yuichi Takano, Kensei Nosaka and Tomoka Takei.

Collaboration pattern

Collaboration is compact: OpenAlex records 5 distinct co-authors in the last three years, with Yuichi Takano, Kensei Nosaka and Tomoka Takei the most frequent partners — the collaboration score of 36 reflects this small network. In both cataloged 2026 papers Ikeda sits in a middle authorship position and is not the corresponding author, indicating a contributing role within a small, stable team rather than a corresponding-author lead.

Research keywords

decision-focused learningpredict-then-optimizeportfolio optimizationmean-variance optimizationbilevel optimizationKKT conditionsstratified samplingonline controlled experimentsA/B testingvariance reductionoptimal decision treesbinary optimization

Highlights

KKT-based decision-focused learning for constrained mean-variance portfolios
A 2026 arXiv paper reformulates constrained mean-variance portfolio optimization as a single-level problem via KKT conditions — claimed as the first such decision-focused learning formulation — with an ETF backtest reporting the best results on SR/FW/CDL metrics; Evolve notes the absence of transaction costs, error bars and significance tests. Repro readiness: medium.
Exact multi-way decision trees for stratified sampling in A/B tests
A second 2026 arXiv paper casts multi-way stratification tree construction for online controlled experiments as an exact binary-optimization problem to cut A/B testing variance; results are promising on two offline datasets, though Evolve notes no field test yet. Repro readiness: high.
Marked output acceleration since 2023
OpenAlex yearly counts rise from 2 indexed works in 2023 to 9 in 2024, dip to 2 in 2025, and reach a record 11 in 2026 (22 works over the last three years); both Evolve-cataloged arXiv preprints appear in 2026.

Timeline

2017First year with an indexed work in the OpenAlex record (1 work).
2018Two indexed works appear, the highest annual count of this early period.
2020One indexed work, after which the record stays quiet until 2023.
2023Activity re-emerges with 2 indexed works, the start of a sustained climb.
2024Output jumps to 9 indexed works, the sharpest single-year rise in the record.
2026A record 11 indexed works, following a dip to 2 in 2025; both Evolve-cataloged arXiv preprints (stratified sampling trees; KKT-based decision-focused learning) appear this year.

Papers in the Evolve catalog

Frequent co-authors

Yuichi Takano×2 Kensei Nosaka×1 Shunnosuke Ikeda×1 Yuichi Takano×1 Tomoka Takei×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.

Bibliometrics are drawn from OpenAlex only; beyond the reported h-index 4 and i10-index 1 there is no usable citation graph, so the output and impact axes remain unscored.
This profile is name-resolved from public bibliographic data; the raw frequent-coauthor list duplicates the researcher's own name and repeats Yuichi Takano, so entity resolution warrants caution.
Both reviewed papers are 2026 arXiv preprints with middle authorship and no corresponding-author role; peer-review status is not covered by the available data.
Yearly counts come from OpenAlex countsByYear and may lag actual publication dates; absolute numbers are small, consistent with an early-stage profile.