Tensor decomposition for multi-agent predictive state representation

作者：

Highlights：

• Propose a tensor-based predictive state representation (PSR).

• Develop the new multi-agent PSR model based on two tensor decompositions methods.

• Use a system dynamics tensor for learning a multi-agent PSR model.

• Compare the new model with state-of-the-art algorithms on multiple problem domains.

摘要

•Propose a tensor-based predictive state representation (PSR).•Develop the new multi-agent PSR model based on two tensor decompositions methods.•Use a system dynamics tensor for learning a multi-agent PSR model.•Compare the new model with state-of-the-art algorithms on multiple problem domains.

论文关键词：Predictive state representations,Tensor optimization,Learning approaches

论文评审过程：Received 9 May 2020, Revised 2 September 2021, Accepted 21 September 2021, Available online 14 October 2021, Version of Record 1 November 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115969