PROJECT TITLE :

Efficient System Tracking With Decomposable Graph-Structured Inputs and Application to Adaptive Equalization With Cyclostationary Inputs - 2018

ABSTRACT:

This Project introduces the graph-structured recursive least squares (GS-RLS) algorithm, which could be a terribly economical suggests that to trace a linear time-varying system when the inputs to the system have structure that can be modeled employing a decomposable Gaussian graphical model. For graphs with small clique sizes, it's shown that GS-RLS will achieve tracking performance terribly close to that of the standard RLS algorithm for a fraction of the computational value. In particular, once proving that the outputs of wide-sense stationary time-varying communication channels have graphical model structure if the inputs are cyclostationary, important computational gains are realized for adaptive equalization of the time-varying underwater acoustic communication channel using the GS-RLS algorithm. This is often verified using field data from the SPACE08 underwater acoustic communication experiment.