LIME

As explained in Ensemble Methods for Machine Learning

LIME is a model-agnostic explanation technique that approximates individual predictions with an interpretable model fitted near the input being explained.

LIME explains one prediction at a time by approximating a complex black-box model with an interpretable linear model near the example of interest. It is model-agnostic, so the black box can be an ensemble or another machine-learning model. section 9.0 section 9.4.1

The procedure follows a local, weighted approximation:

• Start with the example whose prediction needs explanation and generate nearby surrogate examples by perturbing its features. section 9.4.1
• Query the black-box model on those examples, giving nearby examples greater weight than distant ones. section 9.4.1
• Fit a sparse or otherwise interpretable linear surrogate by minimizing a weighted loss between black-box and surrogate predictions. section 9.4.1
• Read the linear coefficients as local feature contributions, while recognizing that the explanation describes a neighborhood rather than the whole model. section 9.4.1 section 9.6

Explaining a subscription prediction

For an XGBoost marketing model, LimeTabularExplainer can be configured with feature names and categorical features, then applied to a test example. Positive contributions support “will subscribe,” while negative contributions support “won’t subscribe.” section 9.4.1

Figure 9.14 LIME creates a surrogate training set of examples in the locality of the example whose prediction needs to be explained. These examples are further weighted by their distance. This is indicated by the sizes of the surrogate examples, with closer examples getting higher weights (and shown larger). A weighted loss function is used to fit a linear surrogate model, which provides local explanations.

Visualize LIME as a weighted cloud of perturbed examples surrounding the target input, with closer examples exerting greater influence on the fitted linear surrogate. section 9.4.1

How Ensemble Methods for Machine Learning covers it

How deep it goes

Defines perturbation, distance weighting, weighted loss, and linear surrogates, but stops short of explaining feature interactions.

Examples

Uses data-driven marketing, an XGBoost subscription model, categorical inputs, and a true-positive test prediction.

What it assumes

Introduces black-box and glass-box models before applying local explanations to individual predictions.

What you can do after

Apply it via lime_tabular to inspect feature contributions for a selected prediction.

How it connects

Connects local surrogate models with linear models, global surrogate models, SHAP, and feature importance.

liveBookmanning.com

Read Ensemble Methods for Machine Learning on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Machine Learning System DesignLIME is a model-agnostic explanation technique that approximates a complex model locally around a particular input with a simpler interpretable model.
• 

  liveBookmanning.com
  Financial AI in PracticeA model-agnostic explanation method that approximates a model locally with an interpretable model around a prediction.

Compare how 3 books explain LIME →

Related concepts

• model-agnostic methodkind of · A model-agnostic method analyzes predictions or estimates explanations without relying on a model’s internal structure, implementation, or architecture.
• local surrogate methodkind of · A local surrogate method explains a black-box model near a particular input by fitting an interpretable approximation to its local behavior.
• interpretabilityused by · Interpretability is the degree to which people can understand a model’s behavior, outputs, predictions, or the factors that produce them.
• local explanationused by · A local explanation describes how or why a model produced a particular prediction or decision for an individual input.
• explainability frameworkused by · An explainability framework is a collection of methods, controls, or tools that helps people understand a model’s outputs or decision process.

This explainer was generated by Manning's knowledge graph from the text of Ensemble Methods for Machine Learning. The section links point to the parts of the book it draws on.