Examples

Here you can find various examples of using FAT Forensics.

Note

Please note that all of the examples using the fatf.utils.models.models.KNN model class may be slow as our implementation of the k-nearest neighbours algorithm is in pure Python, hence is rather slow.

Accountability Examples

Here you can find various examples of how to use FAT Forensics to evaluate accountability – security, safety, robustness and privacy – of data sets, machine learning models and their predictions.

Measuring Robustness of a Data Set – Sampling Bias

Measuring Robustness of a Predictive Model – Systematic Performance Bias

Measuring Robustness of a Prediction – Data Density

Fairness Examples

Here you can find various examples of how to use FAT Forensics to evaluate and mitigate a range of fairness aspects of data sets, machine learning models and their predictions.

Measuring Fairness of a Predictive Model – Disparate Impact

Measuring Fairness of a Prediction – Counterfactual Fairness

Measuring Fairness of a Data Set

Transparency Examples

Here you can find various examples of how to use FAT Forensics to explain and interpret data sets, machine learning models and their predictions.

Using a Surrogate Tree Explainer

Using LIME Explainer

Using Partial Dependence Explainer

Using LIME Image Explainer

Using Data Description Explainer

Using Counterfactual Prediction Explainer

Using Individual Conditional Expectation Explainer