{"title":"On Hyperparameter Determination for GPR-Based Channel Prediction in IRS-Assisted Wireless Communication Systems","authors":"Norisato Suga;Kazuto Yano;Yafei Hou;Toshikazu Sakano","doi":"10.23919/comex.2024XBL0058","DOIUrl":null,"url":null,"abstract":"Intelligent reflecting surface (IRS), which can reflect radio waves controlling the phase of incident radio waves, is being investigated for wireless communication in high-frequency bands. To control the reflection characteristic, it is necessary to separately estimate a large number of channel coefficients between transmitting and receiving antennas through each IRS element. This causes significant overhead for the channel estimation. We have proposed a channel prediction method to reduce the overhead using Gaussian process regression with spectral mixture kernel. In Gaussian process regression, the determination of the hyperparameters used to calculate the kernel matrix has a significant impact on prediction accuracy. In this study, we propose validation-based hyperparameter determination for GPR-based channel prediction and evaluate the performance difference between the gradient method and validation.","PeriodicalId":54101,"journal":{"name":"IEICE Communications Express","volume":null,"pages":null},"PeriodicalIF":0.3000,"publicationDate":"2024-06-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10554672","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEICE Communications Express","FirstCategoryId":"1085","ListUrlMain":"https://ieeexplore.ieee.org/document/10554672/","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 0
Abstract
Intelligent reflecting surface (IRS), which can reflect radio waves controlling the phase of incident radio waves, is being investigated for wireless communication in high-frequency bands. To control the reflection characteristic, it is necessary to separately estimate a large number of channel coefficients between transmitting and receiving antennas through each IRS element. This causes significant overhead for the channel estimation. We have proposed a channel prediction method to reduce the overhead using Gaussian process regression with spectral mixture kernel. In Gaussian process regression, the determination of the hyperparameters used to calculate the kernel matrix has a significant impact on prediction accuracy. In this study, we propose validation-based hyperparameter determination for GPR-based channel prediction and evaluate the performance difference between the gradient method and validation.