Skip to main content

> ML_LIBRARY // LIME_v1.0

LIME

Marco Tulio Ribeiro / University of Washington — Local Interpretable Model-agnostic Explanations for explaining individual black-box predictions.

privacy-security-optimizationv0.2.0.1BSD-2-Clausequalified
Official DocumentationSource Repository

Model Training

Not Supported

This library is a dedicated runtime engine for inference serving and does not train models.

Model Inference

Supported
Inference Accelerators:
CPU
Deployment Targets:server

What It Does

  • +Local surrogate explainability approximating complex black-box models locally with interpretable linear models
  • +Tabular explainer with categorical discretization and feature weighting
  • +Text explainer highlighting positive/negative individual words driving document classification
  • +Image explainer highlighting superpixel regions driving computer vision predictions

What It Does Not Do

  • -Guarantee mathematical consistency or additivity axioms (unlike SHAP)
  • -Provide global dataset-wide explanations (strictly local per-instance explanations)
  • -Run on edge microcontrollers

>Suitable Work Types

  • Explaining why a specific customer support email was classified as "high urgency"
  • Explaining which visual superpixels caused a neural network to classify an image as a medical anomaly
  • Sanity-checking individual predictions for obvious model flaws (e.g. classifying snow as a wolf)

>Unsuitable Work Types

  • Global feature importance across an entire database (use SHAP TreeSHAP)
  • Regulatory compliance requiring game-theoretic mathematical consistency
Data Residency Implications

Runs strictly locally in memory. Zero external calls.

Security Considerations

Permissive BSD-2-Clause license.

Operational Profile & Known Limitations

Maturity:mature
Learning Curve:low
Ops Complexity:low
Cost Tier:free-oss
> Known Limitations:
  • Random sampling around the instance can produce slight instability in feature weights across multiple runs on identical inputs.

Associated Incident Patterns (Incidentpedia)

Enforce safeguards and monitoring to guard against these documented real-world failure modes:

thread-pool-exhaustion

> Primary Evidence & Benchmark Citations

LIME Documentationofficial-docs • 0.2.0.1
2026-09-25HIGH