Swish-Driven GoogleNet for Intelligent Analog Beam Selection in Terahertz Beamspace MIMO

TL;DR

This paper introduces a Swish-driven GoogleNet-based approach for intelligent analog beam selection in terahertz MIMO systems, significantly improving accuracy and spectral efficiency over existing methods.

Contribution

It proposes a novel Swish-activated GoogleNet classifier and an ensemble method for enhanced beam selection in THz MIMO, outperforming traditional techniques.

Findings

Achieves 86% accuracy with Swish activation

Yields 18% spectral efficiency gain over conventional methods

Ensemble classifier reaches 90% accuracy and 27% SE improvement

Abstract

In this paper, we propose an intelligent analog beam selection strategy in a terahertz (THz) band beamspace multiple-input multiple-output (MIMO) system. First inspired by transfer learning, we fine-tune the pre-trained off-the-shelf GoogleNet classifier, to learn analog beam selection as a multi-class mapping problem. Simulation results show 83% accuracy for the analog beam selection, which subsequently results in 12% spectral efficiency (SE) gain, upon the existing counterparts. Towards a more accurate classifier, we replace the conventional rectified linear unit (ReLU) activation function of the GoogleNet with the recently proposed Swish and retrain the fine-tuned GoogleNet to learn analog beam selection. It is numerically indicated that the fine-tuned Swish-driven GoogleNet achieves 86% accuracy, as well as 18% improvement in achievable SE, upon the similar schemes. Eventually, a strong ensembled classifier is developed to learn analog beam selection by sequentially training multiple fine-tuned Swish-driven GoogleNet classifiers. According to the simulations, the strong ensembled model is 90% accurate and yields 27% gain in achievable SE, in comparison with prior methods.