Exponentiated Gradient Reduction

Fairness Transparency Reliability

Description

An in-processing fairness technique based on Agarwal et al.'s reductions approach that transforms fair classification into a sequence of cost-sensitive classification problems. The method uses an exponentiated gradient algorithm to iteratively reweight training data, returning a randomised classifier that achieves the lowest empirical error whilst satisfying fairness constraints. This reduction-based framework provides theoretical guarantees about both accuracy and constraint violation, making it suitable for various fairness criteria including demographic parity and equalised odds.

Example Use Cases

Fairness

Training a hiring algorithm with demographic parity constraints to ensure equal selection rates across gender groups, using iterative reweighting to balance fairness and predictive accuracy whilst maintaining legal compliance.

Transparency

Developing a loan approval model with equalised odds constraints, providing transparent documentation of the theoretical guarantees about both error rates and fairness constraint violations achieved by the reduction approach.

Reliability

Creating a medical diagnosis classifier that maintains reliable performance across demographic groups by using randomised prediction averaging, ensuring consistent healthcare delivery whilst monitoring constraint satisfaction over time.

Limitations

Requires convex base learners for theoretical guarantees about convergence and optimality, limiting the choice of underlying models.
Produces randomised classifiers that may give different predictions for identical inputs, which can be problematic in applications requiring consistent decisions.
Convergence can be slow and sensitive to hyperparameter choices, particularly the learning rate and tolerance settings.
Involves iterative retraining with adjusted weights, which can be computationally expensive for large datasets or complex models.
Fairness constraints may significantly reduce model accuracy, and the trade-off between fairness and performance is not always transparent to practitioners.

Resources

A Reductions Approach to Fair Classification

Research Paper•Alekh Agarwal et al.•Mar 6, 2018

Foundational paper by Agarwal et al. introducing the exponentiated gradient reduction approach for fair classification with theoretical guarantees.

Fairlearn: ExponentiatedGradient

Documentation

Microsoft's Fairlearn implementation of the Agarwal et al. algorithm with comprehensive API documentation and examples.

IBM AIF360: ExponentiatedGradientReduction

Documentation

IBM's AIF360 implementation with scikit-learn compatible API for in-processing fairness constraints during model training.

Fairlearn Reductions Guide

Tutorial

Comprehensive guide to using reduction-based approaches for fairness, including practical examples of exponentiated gradient methods and fairness constraints.

Related Techniques

Name	Description	Assurance Goals
Temperature Scaling	Temperature scaling adjusts a model's confidence by applying a single parameter (temperature) to its predictions. When a model is too confident in its wrong answers, temperature scaling can fix this by making the predictions more realistic. It works by dividing the model's outputs by the temperature value before converting them to probabilities. Higher temperatures make the model less confident, whilst lower temperatures increase confidence. The technique maintains the model's accuracy whilst ensuring that when it says it's 90% confident, it's actually right about 90% of the time.	Reliability Transparency Fairness
ANCHOR	ANCHOR generates high-precision if-then rules that explain individual predictions by identifying the minimal set of feature conditions that guarantee a specific prediction with high confidence. It searches for 'anchor' conditions (e.g., 'age > 30 AND income < £50k') that ensure the model gives the same prediction at least 95% of the time when those conditions are met. This creates human-readable rules that users can trust as sufficient conditions for understanding why a particular decision was made.	Explainability Transparency
Permutation Importance	Permutation Importance quantifies a feature's contribution to a model's performance by randomly shuffling its values and measuring the resulting drop in predictive accuracy. If shuffling a feature significantly degrades the model's performance, that feature is considered important. This model-agnostic technique helps identify which inputs are genuinely driving predictions, rather than just being correlated with the outcome.	Explainability Reliability
Bootstrapping	Bootstrapping estimates uncertainty by repeatedly resampling the original dataset with replacement to create many new training sets, training a model on each sample, and analysing the variation in predictions. This approach provides confidence intervals and stability measures without making strong statistical assumptions. By showing how predictions change with different random samples of the data, it reveals how sensitive the model is to the specific training examples and provides robust uncertainty estimates.	Reliability Transparency Fairness
Empirical Calibration	Empirical calibration adjusts a model's predicted probabilities to match observed frequencies. For example, if events predicted with 80% confidence only occur 60% of the time, calibration would correct this overconfidence. Common techniques include Platt scaling and isotonic regression, which learn transformations that map the model's raw scores to well-calibrated probabilities, improving the reliability of confidence measures for downstream decisions.	Reliability Transparency Fairness
Monte Carlo Dropout	Monte Carlo Dropout estimates prediction uncertainty by applying dropout (randomly setting neural network weights to zero) during inference rather than just training. It performs multiple forward passes through the network with different random dropout patterns and collects the resulting predictions to form a distribution. Low variance across predictions indicates epistemic certainty (the model is confident), while high variance suggests epistemic uncertainty (the model is unsure). This technique transforms any dropout-trained neural network into a Bayesian approximation for uncertainty quantification.	Explainability Reliability

Tags

Applicable Models:

Assurance Goal Category:

Data Requirements:

Sensitive Attributes

Data Type:

Evidence Type:

Quantitative Metric

Fairness Metric

Expertise Needed:

Fairness Approach:

Lifecycle Stage:

Technique Type: