Interactive Demos

• fairlib: A Unified Framework for Assessing and Improving Fairness
  • 1. Installation
  • 2. Build a dataset
  • 3. Train a vanilla model without debiasing
  • 4. Improve Fairness
  • 5. Analyze the results
  • 6. Cutomize pipeline for fairness
• Customized models and datasets for CV
  • Installation
  • Explore Datasets
  • Customizing NN Architectures
  • Customizing Dataloader
  • Training vanilla model without debiasing
  • Improveing Fairness
• Customized models and datasets for structured inputs
  • Installation
  • Download and preprocess the COMPAS dataset
  • Train Models
  • Customize dataset loader
• Load experimental results
  • Basic Plot
  • Zoomed Plots
  • AUC - Performance-Fairness Tradeoff
• Interactive plots
  • Load experimental results
  • Crete Plot
• Manilulating Data Distribution
  • Load data
  • Analysis of the loaded dataset distribution
  • Resample instances based on their target labels and protected labels
  • Limitation and Extension
• fairlib Evaluation Tutorial
  • 1. Installation
  • Train a Model
  • Scenario 1: Confusion Matrix Based Metrics