Olga Fink

—

Unclaimed · AI-structured profilename-matched identity

Machine learning for time series anomaly detection and industrial fault diagnosis: multimodal and graph-based methods

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—

This axis is not scored: the snapshot contains no publication-count or productivity metrics beyond the 2 Evolve-reviewed preprints. It would measure total research output, venue breadth, and volume of work over time, none of which is available in the current input data.

Impact—

This axis is not scored: no citation counts, h-index, or downstream-adoption signals are present in the data. It would measure how widely the two 2026 preprints are cited and taken up by other work, which cannot be estimated from the current snapshot.

Focus0/100

The focus axis scores 0 because this profile's output splits exactly evenly between two fields: machine learning for time series anomaly detection (0.5) and machine learning for industrial fault diagnosis and condition monitoring (0.5). The two are adjacent — both catalog papers target anomaly or fault detection in time series — but with no single subfield dominating the visible record, topical concentration cannot be credited.

Collaboration39/100

The collaboration axis scores 39, based on 6 distinct coauthors in the snapshot. Both reviewed papers carry Olga Fink in the last author position, consistent with a senior or supervising role on small teams rather than large consortia. Recurring collaborators include Zepeng Zhang, Fuad Khuri, Keivan Faghih Niresi, Magnus Munk Jensen, Dorte Hammershøi and Rafał Wiśniewski. The moderate score reflects a compact, stable co-author pool around a focused methodological agenda.

Activity—

This axis is not scored: the snapshot lacks publication-frequency or recency measures beyond the fact that both catalog entries are dated 2026. It would measure publishing cadence and continued research engagement over time, which cannot be assessed from two same-year preprints.

Reproducibility55/100

The reproducibility axis scores 55, the mean readiness across the 2 evaluated papers, both rated medium. The multimodal fault-detection paper reports solid out-of-distribution gains on 3 public datasets but uses thin baselines; GSLAD reports strong gains on 4 benchmarks yet releases no code and does not report variance. The mid-range score reflects public data and clear protocols offset by missing code and statistical reporting.

Research summary

Olga Fink's visible research record in this profile rests on two 2026 arXiv preprints reviewed by Evolve, both with Fink in the last author position — a senior or supervising role — and neither as corresponding author. The work spans two evenly weighted fields: machine learning for time series anomaly detection and machine learning for industrial fault diagnosis and condition monitoring (field distribution 0.5/0.5). The first paper, GSLAD, introduces prototype-regularized graph structure learning for multivariate time series anomaly detection, combining graph neural networks, structural prototypes, and uncertainty-aware distance; Evolve notes strong gains on 4 benchmarks, tempered by missing code and unreported variance. The second proposes self-supervised cross-modal reconstruction for robust fault detection in mechanical multimodal time series with adaptive test-time thresholding; it delivers solid out-of-distribution gains on 3 public datasets, though baselines are thin. Quantitatively, the radar shows a collaboration score of 39 based on 6 distinct coauthors, a focus score of 0 reflecting the even two-field split, and a reproducibility score of 55 averaged over 2 papers, both rated medium readiness. Output, impact, and activity remain unscored because the underlying bibliometric data is unavailable. Frequent collaborators include Zepeng Zhang, Fuad Khuri, Keivan Faghih Niresi, Magnus Munk Jensen, Dorte Hammershøi, and Rafał Wiśniewski. Institution, country, and ORCID are not part of the input data for this profile.

Collaboration pattern

Collaboration in the snapshot is compact: 6 distinct coauthors and a collaboration score of 39. Olga Fink appears as last author on both reviewed 2026 papers — a senior, supervising position rather than hands-on first-author work — with no corresponding-author entries in the catalog. Recurring collaborators include Zepeng Zhang, Fuad Khuri, Keivan Faghih Niresi, Magnus Munk Jensen, Dorte Hammershøi, and Rafał Wiśniewski, suggesting small method-oriented teams rather than large consortia.

Research keywords

multivariate time seriesanomaly detectiongraph structure learninggraph neural networksstructural prototypesuncertainty-aware distancemultimodal learningcross-modal reconstructionfault detectiondistribution shifttest-time adaptationunsupervised anomaly detection

Highlights

GSLAD: prototype-regularized graph structure learning for multivariate time series anomaly detection
In a 2026 arXiv preprint with Olga Fink as last author, GSLAD combines graph structure learning, structural prototypes, and uncertainty-aware distance for multivariate time series anomaly detection. Evolve's review notes strong gains on 4 benchmarks, but the absence of released code and unreported variance temper confidence; reproducibility readiness is rated medium.
Self-supervised cross-modal reconstruction for robust multimodal fault detection
This 2026 arXiv preprint, again with Fink as last author, uses self-supervised cross-modal reconstruction plus adaptive test-time thresholding for fault detection in mechanical multimodal time series. Evolve records solid out-of-distribution gains on 3 public datasets under distribution shift, while noting that baselines are thin; reproducibility readiness is medium.

Timeline

2026Releases GSLAD, an arXiv preprint on prototype-regularized graph structure learning for multivariate time series anomaly detection, in the last-author position.
2026Releases an arXiv preprint on robust fault detection in mechanical multimodal time series via self-supervised cross-modal reconstruction, also as last author.
2026Both preprints enter the Evolve-reviewed catalog with medium reproducibility readiness, giving a mean reproducibility score of 55 across 2 papers.

Papers in the Evolve catalog

Frequent co-authors

Zepeng Zhang×1 Fuad Khuri×1 Keivan Faghih Niresi×1 Magnus Munk Jensen×1 Dorte Hammershøi×1 Rafał Wiśniewski×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.

All bibliometric scores come from a thin snapshot: only 2 Evolve-reviewed arXiv preprints (2026); no journal, conference, or citation data is included.
Output, impact, and activity are unscored because no publication counts, citation counts, or cadence metrics exist in the input; they should not be inferred.
Institution, country, and ORCID are absent from the input; this profile is name-resolved from public bibliographic data, and 'Olga Fink' may not be uniquely disambiguated.
The reproducibility mean of 55 is averaged over only 2 papers, so the score is fragile and may shift as more work is reviewed.