Tensor-network-based predistorter design for multiple-input multiple-output nonlinear systems

Abstract

The recent development in tensor algorithms has shed new light on many design problems previously doomed by the curse of dimensionality. In particular, latest advances are seen in the tensor-network-based multiple-input multiple-output (MIMO) Volterra series modeling of nonlinear systems whereby the Volterra kernels can now be efficiently identified at an unprecedented order and memory length. Subsequent to nonlinear system identification, this paper studies the nonlinear MIMO predistorter design that is crucial for linearizing the response of nonlinear modules such as power amplifiers in mixed-signal applications. Two tensor-network-based predistorter design schemes are presented for the first time, whose effectiveness are validated through practical examples.