University of Florida researcher applying deep learning to error-bounded compression of scientific and video data
2
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
2
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—
No output score yet: the pipeline lacks a complete, deduplicated publication record for this author (the OpenAlex snapshot offers only sparse per-year counts, including a likely-homonym 2001 entry). This axis would measure total publication volume and career-long output growth.
Impact—
Impact is unscored because the citation record is too thin to be meaningful: the snapshot reports h-index 2 and i10-index 0. This axis would measure citation-based influence — how widely the compression methods are adopted and built upon by other groups.
Focus100/100
A perfect focus score of 100 matches the data: of the 18 topic mentions in the OpenAlex profile, 15 sit in compression — Advanced Data Compression Techniques (6), Video Coding and Compression Technologies (5), Algorithms and Data Compression (4) — and the field distribution lists a single field, scientific data compression with machine learning. Only minor side signals (GAN-based image synthesis, 2; plant stress responses, 1) break the pattern.
Collaboration28/100
The collaboration score of 28 reflects a deliberately small working circle: 8 works in the last three years involve only 3 distinct coauthors, and the recurring partners are Sanjay Ranka, Anand Rangarajan, and Surya Majumder. Authorship roles are mixed — first author on the multivariate-compression preprint, middle author on the reviewed RVQ paper, and not corresponding on either — consistent with a junior contributor inside a compact, stable team rather than a broad international network.
Activity64/100
Activity scores 64 on the strength of a clear recent uptick: the OpenAlex countsByYear series shows 4 works in 2025 and 4 in 2026 — 8 works within the recent three-year window — after a single indexed work in 2023. Both 2026 arXiv preprints in the Evolve catalog fall inside this burst, suggesting compression research is the current primary effort rather than an occasional side interest.
Reproducibility55/100
The reproducibility score of 55 (reproMean) rests on a single assessed paper, so confidence is limited. For the RVQ + pixel-U-Net + Guaranteed Autoencoder compression pipeline, Evolve set reproReadiness to medium: quantitative claims appear largely as figures, classical compression baselines are absent, and the error-bound guarantee is thin — code, per-point numbers, and baseline tables would be needed before independent replication.
Research summary
Liangji Zhu is an early-career researcher at the University of Florida working where machine learning meets scientific data compression. The OpenAlex snapshot is thin but coherent: h-index 2, i10-index 0, and 8 works in the last three years, concentrated in 2025 (4 works) and 2026 (4 works). Topic clustering is unusually tight — Advanced Data Compression Techniques (6 mentions), Video Coding and Compression Technologies (5), and Algorithms and Data Compression (4) — with only minor side signals from GAN-based image synthesis (2) and plant stress responses (1, possibly a homonym artifact). Two 2026 arXiv preprints define the current agenda. The Evolve-reviewed paper proposes a residual-vector-quantization pipeline with a pixel-space U-Net and a Guaranteed Autoencoder for error-bounded lossy compression, reporting lower NRMSE at equal compression ratio across three datasets; Evolve judges it promising but flags figure-only results, missing classical baselines, and thin guarantees, with medium reproducibility readiness. The second, first-authored preprint extends learned compression to multivariate scientific data via cross-variable latent decorrelation and autoregressive entropy modeling. Co-authorship stays within a small stable circle — Sanjay Ranka, Anand Rangarajan, and Surya Majumder — making this a focused, early-stage profile whose citation record has not yet caught up with the recent output.
Collaboration pattern
Collaboration is narrow and stable: 8 works over the last three years with only 3 distinct coauthors, and the same names recur — Sanjay Ranka, Anand Rangarajan, and Surya Majumder. Roles alternate between first author (the multivariate-compression preprint) and middle author (the reviewed RVQ paper), with no corresponding-author credit in the catalog, consistent with a junior researcher embedded in a small University of Florida team rather than a large distributed consortium.
Research keywords
Highlights
Error-bounded neural compression with residual vector quantization
First-author push into multivariate scientific data compression
A second 2026 arXiv preprint, with Zhu as first author, targets multivariate scientific data through learned cross-variable latent decorrelation and autoregressive entropy modeling; it is newly in the catalog and not yet reviewed by Evolve, so no verdict or reproducibility rating exists yet.
A tightly focused compression research program
OpenAlex topic counts show 15 of 18 topic mentions in compression — advanced data compression (6), video coding (5), algorithms and data compression (4) — against minor GAN/image-synthesis (2) and plant-stress (1) signals, yielding the maximum focus score of 100 in the Evolve radar.
Timeline
- 2023Earliest recent entry in the OpenAlex countsByYear series: 1 indexed work, marking the visible start of the publication record in the snapshot.
- 2025Output steps up to 4 indexed works, the first sustained burst of activity in the snapshot.
- 20264 indexed works, including two arXiv preprints on neural scientific data compression — one first-authored (multivariate latent decorrelation) and one Evolve-reviewed (residual vector quantization with pixel-space U-Net and Guaranteed Autoencoder).
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.
- All bibliometrics come from a single OpenAlex snapshot; Google Scholar or Scopus coverage may differ.
- The profile may aggregate homonym records: the 2001 countsByYear entry and the plant-stress topic are unrelated to the compression agenda.
- Output and impact dimensions are unscored because the underlying publication and citation records are insufficient.
- Only one of the two catalog papers has been reviewed by Evolve; the multivariate-compression preprint has no verdict, keywords, or slug yet.