Modeling Massive Amount of Experimental Data with Large Random Matrices in a Real-Time UWB-MIMO System

Abstract

The aim of this paper is to study data modeling for massive datasets. Large random matrices are used to model the massive amount of data collected from our experimental testbed. This testbed was developed for a real-time ultra-wideband, multiple input multiple output (UWB-MIMO) system. Empirical spectral density is the relevant information we seek for. After we treat this UWB-MIMO system as a black box, we aim to model the output of the black box as a large statistical system, whose outputs can be described by (large) random matrices. This model is extremely general to allow for the study of non-linear and non-Gaussian phenomenon. The good agreements between the theoretical predictions and the empirical findings validate the correctness of the our suggested data model.