🔍 LIME (Local Interpretable Model-agnostic Explanations)

LIME (Local Interpretable Model-agnostic Explanations)#

Explaining a prediction by fitting a simple local surrogate model.

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

What it is#

LIME (Local Interpretable Model-agnostic Explanations) explains an individual prediction of any ML model by approximating the black box locally — around the instance of interest — with a simpler, interpretable model such as linear regression. It answers: why did the model predict this for this example?

How it works#

Six steps. Take the instance to explain; create perturbed samples by slightly varying its features; collect the black-box predictions for those samples; weight each sample by proximity to the original; fit a simple interpretable model (linear or tree) on that local neighbourhood; and read its coefficients as the feature contributions.

A worked example#

For a denied loan, perturbing income, age and debt and fitting a local linear model might yield Income -0.4, high debt +0.3 and employment length +0.1 — low income plus high debt pushed the decision toward denial.

Strengths, limits, and SHAP#

LIME is model-agnostic, sharply local (one prediction at a time) and human-friendly, but it is unstable (different perturbations give different explanations), only locally faithful, computationally expensive, and shaky under correlated features. Against SHAP: LIME fits local surrogate models (faster, less stable) while SHAP uses game theory (local and global, more stable, an exact decomposition).

In practice#

import lime.lime_tabular
from sklearn.datasets import load_iris
from sklearn.ensemble import RandomForestClassifier

X, y = load_iris(return_X_y=True)
model = RandomForestClassifier().fit(X, y)

explainer = lime.lime_tabular.LimeTabularExplainer(
    X,
    feature_names=["f1", "f2", "f3", "f4"],
    class_names=["setosa", "versicolor", "virginica"],
    discretize_continuous=True,
)
exp = explainer.explain_instance(X[0], model.predict_proba, num_features=2)
exp.show_in_notebook()

Theme: Explainability & Governance · All terminology

Hint

Mind map — connected ideas

SHAP (SHapley Additive exPlanations) · Counterfactual Explanations · Post-hoc Explainability · Feature Values · Discriminatory Power · Deep Ensembles

Hint

More in Explainability & Governance

Basel III · Counterfactual Explanations · Fair Lending laws · High-Stakes Domains · Post-hoc Explainability · SHAP (SHapley Additive exPlanations)