Model-Based Reinforcement Learning via Imagination with Derived Memory

no code implementations • NeurIPS 2021 • Yao Mu, Yuzheng Zhuang, Bin Wang, Guangxiang Zhu, Wulong Liu, Jianyu Chen, Ping Luo, Shengbo Li, Chongjie Zhang, Jianye Hao

Model-based reinforcement learning aims to improve the sample efficiency of policy learning by modeling the dynamics of the environment.

Model-based Reinforcement Learning reinforcement-learning +2

Offline Reinforcement Learning with Reverse Model-based Imagination

1 code implementation • NeurIPS 2021 • Jianhao Wang, Wenzhe Li, Haozhe Jiang, Guangxiang Zhu, Siyuan Li, Chongjie Zhang

These reverse imaginations provide informed data augmentation for model-free policy learning and enable conservative generalization beyond the offline dataset.

Data Augmentation Offline RL +3

On the Estimation Bias in Double Q-Learning

1 code implementation • NeurIPS 2021 • Zhizhou Ren, Guangxiang Zhu, Hao Hu, Beining Han, Jianglun Chen, Chongjie Zhang

Double Q-learning is a classical method for reducing overestimation bias, which is caused by taking maximum estimated values in the Bellman operation.

Q-Learning Value prediction

Generalizable Episodic Memory for Deep Reinforcement Learning

1 code implementation • 11 Mar 2021 • Hao Hu, Jianing Ye, Guangxiang Zhu, Zhizhou Ren, Chongjie Zhang

Episodic memory-based methods can rapidly latch onto past successful strategies by a non-parametric memory and improve sample efficiency of traditional reinforcement learning.

Atari Games Continuous Control +3

Bridging Imagination and Reality for Model-Based Deep Reinforcement Learning

1 code implementation • NeurIPS 2020 • Guangxiang Zhu, Minghao Zhang, Honglak Lee, Chongjie Zhang

It maximizes the mutual information between imaginary and real trajectories so that the policy improvement learned from imaginary trajectories can be easily generalized to real trajectories.

Model-based Reinforcement Learning reinforcement-learning +2

Object-Oriented Model Learning through Multi-Level Abstraction

no code implementations • ICLR 2019 • Guangxiang Zhu, Jianhao Wang, Zhizhou Ren, Chongjie Zhang

Object-based approaches for learning action-conditioned dynamics has demonstrated promise for generalization and interpretability.

Object Relational Reasoning +1

Object-Oriented Dynamics Learning through Multi-Level Abstraction

1 code implementation • 16 Apr 2019 • Guangxiang Zhu, Jianhao Wang, Zhizhou Ren, Zichuan Lin, Chongjie Zhang

We also design a spatial-temporal relational reasoning mechanism for MAOP to support instance-level dynamics learning and handle partial observability.

Object Relational Reasoning +1

Context-Aware Policy Reuse

no code implementations • 11 Jun 2018 • Siyuan Li, Fangda Gu, Guangxiang Zhu, Chongjie Zhang

Transfer learning can greatly speed up reinforcement learning for a new task by leveraging policies of relevant tasks.

Reinforcement Learning Transfer Learning

Object-Oriented Dynamics Predictor

1 code implementation • NeurIPS 2018 • Guangxiang Zhu, Zhiao Huang, Chongjie Zhang

Generalization has been one of the major challenges for learning dynamics models in model-based reinforcement learning.

Model-based Reinforcement Learning Object +1

Generative Adversarial Mapping Networks

no code implementations • 28 Sep 2017 • Jianbo Guo, Guangxiang Zhu, Jian Li

They fit generative models by minimizing certain distance measure between the real image distribution and the generated data distribution.