Presentation | 2023-03-02 Memory-Driven MLP-Mixer Deep Learning Model for Intelligent Reflecting Surface-Assisted mmWave Systems Beam Prediction Taisei Urakami, Haohui Jia, Na Chen, Minoru Okada, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Although deep learning (DL) based beam prediction for intelligent reflecting surface (IRS)-assisted millimeter wave (mmWave) system is capable of improving the transmission efficiency, it has drawback in limited learning efficiency, robustness for the noise, and large training overhead. In order to address these issues, this paper proposes the memory-driven multi-perceptron-mixer (MDMLP-Mixer) DL model, which consists of the autoregressive encoder (ARE) and spatial MLP-Mixer. Specifically, we use a gated recurrent unit (GRU) as ARE to extract frequency features based on the channel state information (CSI) dataset. In the spatial MLP-Mixer, spatial correlations can be extracted to obtain global features from the frequency and spatial domains by individually mixing the frequency features from each element at IRS index. By the channel matrix compression method the sparsity of CSI, our proposal realizes high beam prediction accuracy and small training overhead, since it achieves a sufficient average accuracy of 80.7% with 10min 52s of simulation time and 14k of parameters. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | mmWave / Intelligent Reflecting Surface / Beam Prediction / Deep Learning |
Paper # | MW2022-159 |
Date of Issue | 2023-02-23 (MW) |
Conference Information | |
Committee | MW |
---|---|
Conference Date | 2023/3/2(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Tottori Univ. |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Microwave, etc. |
Chair | Noriharu Suematsu(Tohoku Univ.) |
Vice Chair | Tadashi Kawai(Univ. of Hyogo) / Kensuke Okubo(Okayama Prefectural Univ.) / Hideyuki Nakamizo(Mitsubishi Electric) |
Secretary | Tadashi Kawai(Univ. of Electro-Comm) / Kensuke Okubo(Toshiba) / Hideyuki Nakamizo |
Assistant | Naoki Hasegawa(Softbank) / Kosuke Katayama(NIT Tokuyama College) |
Paper Information | |
Registration To | Technical Committee on Microwaves |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Memory-Driven MLP-Mixer Deep Learning Model for Intelligent Reflecting Surface-Assisted mmWave Systems Beam Prediction |
Sub Title (in English) | |
Keyword(1) | mmWave |
Keyword(2) | Intelligent Reflecting Surface |
Keyword(3) | Beam Prediction |
Keyword(4) | Deep Learning |
1st Author's Name | Taisei Urakami |
1st Author's Affiliation | Nara Institute of Science and Technology(NAIST) |
2nd Author's Name | Haohui Jia |
2nd Author's Affiliation | Nara Institute of Science and Technology(NAIST) |
3rd Author's Name | Na Chen |
3rd Author's Affiliation | Nara Institute of Science and Technology(NAIST) |
4th Author's Name | Minoru Okada |
4th Author's Affiliation | Nara Institute of Science and Technology(NAIST) |
Date | 2023-03-02 |
Paper # | MW2022-159 |
Volume (vol) | vol.122 |
Number (no) | MW-411 |
Page | pp.pp.13-18(MW), |
#Pages | 6 |
Date of Issue | 2023-02-23 (MW) |