Suboptimal Demonstration

Suboptimal demonstration in machine learning focuses on improving algorithms' ability to learn effectively from imperfect or incomplete training data, a common scenario in real-world applications. Current research emphasizes developing robust algorithms, often incorporating techniques like inverse reinforcement learning, contrastive learning, and actor-critic methods, to filter out noise, prioritize high-quality segments of demonstrations, and learn from diverse data sources including both expert and suboptimal examples. This research area is crucial for advancing the practicality of machine learning, enabling more efficient and reliable training of robots, AI agents, and other systems in situations where perfect data is unavailable or costly to obtain.

Papers

October 14, 2022

Monte Carlo Augmented Actor-Critic for Sparse Reward Deep Reinforcement Learning from Suboptimal Demonstrations
Albert Wilcox, Ashwin Balakrishna, Jules Dedieu, Wyame Benslimane, Daniel S. Brown, Ken Goldberg
Reinforcement Learning Sparse Reward Suboptimal Demonstration

September 26, 2022

Understanding Hindsight Goal Relabeling from a Divergence Minimization Perspective
Lunjun Zhang, Bradly C. Stadie
Imitation Learning Suboptimal Demonstration Hindsight Instruction Multi Goal Reinforcement Learning Divergence Optimization Hindsight Goal Relabeling

July 20, 2022

Discriminator-Weighted Offline Imitation Learning from Suboptimal Demonstrations
Haoran Xu, Xianyuan Zhan, Honglei Yin, Huiling Qin
Offline Reinforcement Learning Reward Learning Suboptimal Demonstration Offline Imitation Learning Expert Policy

July 7, 2022

Wrapping Haptic Displays Around Robot Arms to Communicate Learning
Antonio Alvarez Valdivia, Soheil Habibian, Carly A. Mendenhall, Francesco Fuentes, Ritish Shailly, Dylan P. Losey, Laura H. Blumenschein
Robot Arm Suboptimal Demonstration Visual Feedback Communication Learning Haptic Display

May 23, 2022

Efficient Reinforcement Learning from Demonstration Using Local Ensemble and Reparameterization with Split and Merge of Expert Policies
Yu Wang, Fang Liu
Reinforcement Learning Policy Gradient Split Learning Textual Demonstration Efficient Reinforcement Learning State Action Pair Reparameterization Method Suboptimal Demonstration Expert Policy

May 5, 2022

Semi-Supervised Imitation Learning of Team Policies from Suboptimal Demonstrations
Sangwon Seo, Vaibhav V. Unhelkar
Imitation Learning Suboptimal Demonstration Multi Agent Task Human Agent Team Multi Agent Imitation Conditional Imitation Teammate Policy

March 21, 2022

Self-Imitation Learning from Demonstrations
Georgiy Pshikhachev, Dmitry Ivanov, Vladimir Egorov, Aleksei Shpilman
Reinforcement Learning Sparse Reward Noisy Demonstration Suboptimal Demonstration Learning From Demonstration

November 16, 2021

Improving Learning from Demonstrations by Learning from Experience
Haofeng Liu, Yiwen Chen, Jiayi Tan, Marcelo H Ang
Reinforcement Learning Behavior Cloning Noisy Demonstration Optimal Trajectory eXperience Report Suboptimal Demonstration Enhanced Learning

Suboptimal Demonstration

Papers

Monte Carlo Augmented Actor-Critic for Sparse Reward Deep Reinforcement Learning from Suboptimal Demonstrations

Understanding Hindsight Goal Relabeling from a Divergence Minimization Perspective

Discriminator-Weighted Offline Imitation Learning from Suboptimal Demonstrations

Wrapping Haptic Displays Around Robot Arms to Communicate Learning

Efficient Reinforcement Learning from Demonstration Using Local Ensemble and Reparameterization with Split and Merge of Expert Policies

Semi-Supervised Imitation Learning of Team Policies from Suboptimal Demonstrations

Self-Imitation Learning from Demonstrations

Improving Learning from Demonstrations by Learning from Experience