Tensor optimization with group lasso for multi-agent predictive state representation

作者：

Highlights：

• We propose a tensor optimization method for learning multi-agent PSR model based on ADMM technique. The PSR model parameters can be obtained directly from the matrix X without extra computations.

• We use the original information from system dynamics tensor as core tests and generalize the learning approaches in a multi-agent setting.

• We construct a sparse representation of system dynamics tensor for a multi-agent PSR model and utilize mapping technology for speed up our algorithm.

摘要

•We propose a tensor optimization method for learning multi-agent PSR model based on ADMM technique. The PSR model parameters can be obtained directly from the matrix X without extra computations.•We use the original information from system dynamics tensor as core tests and generalize the learning approaches in a multi-agent setting.•We construct a sparse representation of system dynamics tensor for a multi-agent PSR model and utilize mapping technology for speed up our algorithm.

论文关键词：Predictive state representations,Tensor optimization,Alternating direction method of multipliers,Group lasso

论文评审过程：Received 18 July 2020, Revised 1 January 2021, Accepted 21 February 2021, Available online 17 March 2021, Version of Record 24 March 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.106893