Regret Optimal Algorithm

Regret-optimal algorithms aim to minimize the difference between an algorithm's cumulative reward and that of an optimal strategy with perfect foresight, a crucial problem across various machine learning domains. Current research focuses on developing efficient algorithms for diverse settings, including collaborative bandits with budget constraints, federated learning, and contextual Markov decision processes, often employing techniques like online function approximation and matrix completion. These advancements are significant for improving the performance of online learning systems in applications ranging from recommendation systems and online advertising to reinforcement learning and financial modeling. The pursuit of efficient and provably optimal algorithms continues to drive progress in these areas.

Papers

October 10, 2024

Accurate and Regret-aware Numerical Problem Solver for Tabular Question Answering
Yuxiang Wang, Jianzhong Qi, Junhao Gan
Multi Step Reasoning Answer Generation Answer Correctness Table Question Regret Optimal Algorithm Free Form

November 6, 2023

Online Learning Quantum States with the Logarithmic Loss via VB-FTRL
Wei-Fu Tseng, Kai-Chun Chen, Zi-Hong Xiao, Yen-Huan Li
Quantum State Logarithmic Barrier Logarithmic Loss Regret Optimal Algorithm

October 31, 2023

Blocked Collaborative Bandits: Online Collaborative Filtering with Per-Item Budget Constraints
Soumyabrata Pal, Arun Sai Suggala, Karthikeyan Shanmugam, Prateek Jain
Multi Armed Bandit Collaborative Filtering Sub Linear Regret Budget Constraint Regret Optimal Algorithm Latent Bandit Collaborative Bandit

September 8, 2023

Regret-Optimal Federated Transfer Learning for Kernel Regression with Applications in American Option Pricing
Xuwei Yang, Anastasis Kratsios, Florian Krach, Matheus Grasselli, Aurelien Lucchi
Financial Application Kernel Regression Option Pricing Federated Transfer Learning Training Pair Regret Optimal Algorithm

March 2, 2023

November 16, 2022

Bandit Algorithms for Prophet Inequality and Pandora's Box
Khashayar Gatmiry, Thomas Kesselheim, Sahil Singla, Yifan Wang
Multi Armed Bandit Stochastic Optimization Regret Minimization Object Box Least Absolute Deviation PanDORA Dataset Regret Optimal Algorithm

September 7, 2022

Dual Instrumental Method for Confounded Kernelized Bandits
Xueping Gong, Jiheng Zhang
Instrumental Variable Bandit Problem Contextual Bandit Problem Kernel Bandit Dual Approach Regret Optimal Algorithm

February 13, 2022

Sample-Efficient Reinforcement Learning with loglog(T) Switching Cost
Dan Qiao, Ming Yin, Ming Min, Yu-Xiang Wang
Reinforcement Learning Hidden CoST Sample Efficient Reinforcement Learning Regret Algorithm Adaptive Policy Reward Free Exploration Regret Optimal Algorithm

December 2, 2021

Convergence Guarantees for Deep Epsilon Greedy Policy Learning
Michael Rawson, Radu Balan
Convergence Guarantee Reinforcement Learning Problem Epsilon Greedy Regret Optimal Algorithm

November 24, 2021

Efficient and Optimal Algorithms for Contextual Dueling Bandits under Realizability
Aadirupa Saha, Akshay Krishnamurthy
High Efficiency Regret Minimization Optimal Algorithm Dynamic Regret Dueling Bandit Multiple Realizability Regret Optimal Algorithm

Regret Optimal Algorithm

Papers

Accurate and Regret-aware Numerical Problem Solver for Tabular Question Answering

Online Learning Quantum States with the Logarithmic Loss via VB-FTRL

Blocked Collaborative Bandits: Online Collaborative Filtering with Per-Item Budget Constraints

Regret-Optimal Federated Transfer Learning for Kernel Regression with Applications in American Option Pricing

Efficient Rate Optimal Regret for Adversarial Contextual MDPs Using Online Function Approximation

Learning not to Regret

Bandit Algorithms for Prophet Inequality and Pandora's Box

Dual Instrumental Method for Confounded Kernelized Bandits

Sample-Efficient Reinforcement Learning with loglog(T) Switching Cost

Convergence Guarantees for Deep Epsilon Greedy Policy Learning

Efficient and Optimal Algorithms for Contextual Dueling Bandits under Realizability