Applied deep learning researcher at Seoul National University: weak supervision, GNN quantization, astronomical imaging
3
papers here
0
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
6
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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 data is insufficient to score this axis. Once fuller venue and publication-type records are linked, it would measure annual publication volume, output venues, and authorship roles across the researcher's career.
Impact—
Citation-linked records are too sparse to score impact. This axis would measure citation counts, h-index trajectory, and influence on follow-up work once more complete citation data becomes available.
Focus—
Topic-classification coverage is insufficient to score focus. Once richer subject labels and venue fields are attached to each work, this axis would measure topical concentration versus dispersion across the portfolio.
Collaboration53/100
Scoring 53/100, collaboration is moderate. The OpenAlex record lists 13 distinct coauthors over the last three years alongside 9 works in that window, with recurring partners including Dain Kwon, Jinho Lee, Hyeyoon Lee, Jeong-Eun Lee, and Sunjong Park. All three 2026 catalog papers place this researcher in a middle author position, consistent with contribution to small, domain-specific applied teams rather than solo or corresponding-author leadership.
Activity67/100
Scoring 67/100, activity is comparatively strong. The counts-by-year series shows continuous indexed output from 2015 to 2026, with peaks of 5 works in 2021 and 5 in 2025, and 3 works already indexed in 2026; the last three years contribute 9 of the 21 total indexed works, indicating a sustained and accelerating publication cadence in applied machine learning.
Reproducibility—
No artifact or code-availability assessments exist yet, so reproducibility cannot be scored. This axis would measure code release, dataset availability, and Evolve repro-readiness checks across the reviewed papers.
Research summary
Kanghyun Choi is a researcher at Seoul National University whose work, as indexed by OpenAlex, centers on applied deep learning — advanced neural network applications (11 indexed works), domain adaptation and few-shot learning (7), and parallel computing and optimization (3), with further clusters in adversarial robustness (3) and numerical methods (3). The record spans 2015–2026 with 21 indexed works, an h-index of 6 and an i10-index of 5, including 9 works in the last three years and 13 distinct coauthors, indicating a steadily active, collaboration-driven applied profile. Three 2026 arXiv preprints in the Evolve catalog illustrate the range: SePArate applies weak supervision to segment defect patterns in wafer manufacturing; TopGQ introduces a fast, topology-aware post-training quantization method for graph neural networks; and STARLINC removes satellite-trail artifacts from astronomical images using inter-frame correlation. Together they show a consistent pattern: taking established learning techniques — weak supervision, quantization, temporal correlation — and engineering them for domain-specific measurement and imaging problems, from semiconductor inspection to astronomy. All three cataloged papers list the researcher in a middle author position without corresponding-author designation, consistent with a contributor role on small applied teams. Because citation-linked records, venue classifications, and artifact-availability data are thin in the current snapshot, output, impact, focus, and reproducibility axes remain unscored, and this profile should be read as an early-stage, bibliometrics-grounded view rather than a complete assessment.
Collaboration pattern
The record shows 13 distinct coauthors in the last three years, with a recurring core including Dain Kwon, Sunjong Park, Jinho Lee, Hyeyoon Lee, Shingeon Kim, and Jeong-Eun Lee. All three 2026 catalog papers list middle authorship without corresponding designation, fitting a pattern of contributing specialist within small applied teams spanning semiconductor inspection, graph learning efficiency, and observational astronomy.
Research keywords
Highlights
SePArate: weak supervision for wafer defect pattern segmentation
A 2026 arXiv preprint in the Evolve catalog applying weak supervision to segment recurring defect patterns in wafer manufacturing, targeting inspection settings where full annotation is costly.
TopGQ: topology-aware post-training quantization for GNNs
A 2026 arXiv preprint proposing a fast post-training quantization scheme for graph neural networks that leverages topology information, addressing efficient deployment of GNNs.
STARLINC: removing satellite trails from astronomical images
A 2026 arXiv preprint that removes satellite-trail artifacts from astronomical observations using inter-frame correlation, connecting the researcher's machine-learning work to observational astronomy.
Timeline
- 2015Earliest indexed work appears in the OpenAlex counts-by-year record (1 work), beginning a publication trail that continues through 2026.
- 2021Indexed output reaches an early peak of 5 works in the counts-by-year series.
- 2025Output rebounds to 5 indexed works, opening the most active stretch on record.
- 2026Three arXiv preprints enter the Evolve catalog — SePArate, TopGQ, and STARLINC — with 3 works indexed year-to-date.
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 (h-index, i10-index, yearly counts, topics) come from an OpenAlex-derived snapshot and may be incomplete or affected by same-name author conflation.
- The three reviewed papers are 2026 arXiv preprints; the catalog records no peer-review verdict, keywords, or reproducibility assessment for them.
- Output, impact, focus, and reproducibility dimensions are unscored because the underlying data is insufficient in this snapshot.
- All cataloged papers list middle authorship with no corresponding-author role, so leadership claims cannot be made from this data.