Provable Benefit

Provable benefit research aims to rigorously establish the advantages of various machine learning techniques beyond empirical observation. Current efforts focus on theoretically analyzing the efficiency and effectiveness of methods like annealed Langevin Monte Carlo for sampling, sparse mixture of experts models with perturbed cosine routers, and curriculum learning for neural networks, often within specific problem settings (e.g., high-dimensional sparse Gaussian classification, reinforcement learning). These advancements provide a deeper understanding of algorithm performance, leading to improved model design and more reliable predictions in diverse applications, from reinforcement learning to quantum machine learning.

Papers

February 27, 2023

The Provable Benefits of Unsupervised Data Sharing for Offline Reinforcement Learning
Hao Hu, Yiqin Yang, Qianchuan Zhao, Chongjie Zhang
Offline Reinforcement Learning Offline Reinforcement Learning Algorithm Provable Benefit Reward Free Self Supervised Reinforcement Learning

January 30, 2023

On the Statistical Benefits of Temporal Difference Learning
David Cheikhi, Daniel Russo
Temporal Difference Learning Provable Benefit Crossing Decision Finite State Markov

October 19, 2022

On the Power of Pre-training for Generalization in RL: Provable Benefits and Hardness
Haotian Ye, Xiaoyu Chen, Liwei Wang, Simon S. Du
Reinforcement Learning Strong Generalization Real Power Hardness Result Near Optimal Policy State Action Space Provable Benefit

September 20, 2022

Relational Reasoning via Set Transformers: Provable Efficiency and Applications to MARL
Fengzhuo Zhang, Boyi Liu, Kaixin Wang, Vincent Y. F. Tan, Zhuoran Yang, Zhaoran Wang
Financial Application Relational Reasoning Provable Benefit

June 13, 2022

Provable Benefit of Multitask Representation Learning in Reinforcement Learning
Yuan Cheng, Songtao Feng, Jing Yang, Hong Zhang, Yingbin Liang
Reinforcement Learning Low Rank Provable Benefit Multi Task Representation Reward Free Reward Free Reinforcement Learning

May 29, 2022

Provable Benefits of Representational Transfer in Reinforcement Learning
Alekh Agarwal, Yuda Song, Wen Sun, Kaiwen Wang, Mengdi Wang, Xuezhou Zhang
Reinforcement Learning Transferable Representation Provable Benefit

November 19, 2021

November 13, 2021

On the Statistical Benefits of Curriculum Learning
Ziping Xu, Ambuj Tewari
Curriculum Learning Adaptive Learning Structured Document Provable Benefit Local Prediction

Provable Benefit

Papers

The Provable Benefits of Unsupervised Data Sharing for Offline Reinforcement Learning

On the Statistical Benefits of Temporal Difference Learning

On the Power of Pre-training for Generalization in RL: Provable Benefits and Hardness

Relational Reasoning via Set Transformers: Provable Efficiency and Applications to MARL

Provable Benefit of Multitask Representation Learning in Reinforcement Learning

Provable Benefits of Representational Transfer in Reinforcement Learning

Gaussian Process Inference Using Mini-batch Stochastic Gradient Descent: Convergence Guarantees and Empirical Benefits

DyFormer: A Scalable Dynamic Graph Transformer with Provable Benefits on Generalization Ability

On the Statistical Benefits of Curriculum Learning