Welcome to Dueling Bandit Toolkit’s Documentation!

The Dueling Bandit Toolkit is a Python package for preference-based online learning using dueling bandit algorithms. It implements algorithms like PARWiS, Contextual PARWiS, RL PARWiS, Double Thompson Sampling, and a Random Pair baseline, with support for synthetic and real-world datasets (Jester, MovieLens).

This documentation is based on the research paper PARWiS: Winner determination under shoestring budgets using active pairwise comparisons by Shailendra Bhandari, providing an overview, methodology, experimental results, and API references.

Contents:

• Introduction
  • Overview
  • Background
• Methodology
  • Problem Setting
  • Algorithms
  • Datasets
  • Evaluation Metrics
• Experiments
  • Setup
  • Results
• API Reference
  • Modules
• References

Indices and tables

• Index

• Module Index

• Search Page