Skip to main navigation Skip to search Skip to main content

Checking Unsupervised Learning for Nondeterminism and Inconsistency via SMT Solving

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Unsupervised Learning (UL) implementations are widely used, but prone to issues such as nondeterminism and inconsistency, which makes them unreliable and exploitable. Moreover, the lack of a specification (or ground truth) makes UL implementations hard to verify or test: when they produce diverging results on a dataset, they do so subtly and silently, which gives users a false sense of security, and complicates developers' job of tracing the error to its root cause. We introduce an approach, Ocelot, that reduces nondeterminism/inconsistency checking to satisfiability checking in Z3. Given the Python code of an UL implementation, Ocelot automatically isolates a nondeterministic "kernel"(variables and code), which forms the basis for a Z3 specification. A Z3-found solution acts as a "witness"dataset - a typically small dataset that can be used to reliably reproduce the error. Ocelot has exposed sources of nondeterminism and inconsistency in popular implementations (Scikit-learn, MATLAB, and R) of well-known Clustering algorithms (Affinity Propagation, DBSCAN, HAC, and K-means) and Anomaly Detection algorithms (Isolation Forest, Local Outlier Factor). Moreover, Ocelot has exposed regressions (changes in semantics) in the evolution of these implementations.

Original languageEnglish (US)
Title of host publicationProceedings - 2026 IEEE/ACM 48th International Conference on Software Engineering, ICSE 2026
PublisherAssociation for Computing Machinery, Inc
Pages2975-2987
Number of pages13
ISBN (Electronic)9798400720253
DOIs
StatePublished - Sep 11 2026
Event48th International Conference on Software Engineering, ICSE 2026 - Rio de Janeiro, Brazil
Duration: Apr 12 2026 → Apr 18 2026

Publication series

NameProceedings - 2026 IEEE/ACM 48th International Conference on Software Engineering, ICSE 2026

Conference

Conference48th International Conference on Software Engineering, ICSE 2026
Country/TerritoryBrazil
CityRio de Janeiro
Period4/12/26 → 4/18/26

All Science Journal Classification (ASJC) codes

  • Software
  • Industrial and Manufacturing Engineering
  • Safety, Risk, Reliability and Quality

Keywords

  • Formal checking
  • SMT Solving
  • Unsupervised Learning

Fingerprint

Dive into the research topics of 'Checking Unsupervised Learning for Nondeterminism and Inconsistency via SMT Solving'. Together they form a unique fingerprint.

Cite this