講演抄録/キーワード |
講演名 |
2022-07-15 13:50
Prediction of E-field Distribution in Indoor Environments Using Deep Learning Technique ○Liu Sen・Onishi Teruo・Taki Masao・Watanabe Soichi(NICT) EMCJ2022-34 |
抄録 |
(和) |
(まだ登録されていません) |
(英) |
As one of the important aspects of monitoring electromagnetic field (EMF) exposure levels, comprehensively grasping the electric-field (E-field) distribution in an indoor environment is challenging. Essentially, it belongs to creating a surrogate model in a highly uncertain and variable condition. Benefit from the advancements in deep learning, in this paper, we present two prediction models with one based on fully-connected neural networks (FCNNs) and the other one based on graph neural networks (GNNs). The models are trained upon a same dataset, and are compared to each other. We demonstrate that both models can be used to predict the field distribution in a floorplan geometry that is beyond the training data. We emphasize that GNNs are more powerful than FCNNs due to their non-Euclidean data handling feature. |
キーワード |
(和) |
/ / / / / / / |
(英) |
Deep learning / EMF monitoring / Machine learning / Surrogate model / / / / |
文献情報 |
信学技報, vol. 122, no. 112, EMCJ2022-34, pp. 1-5, 2022年7月. |
資料番号 |
EMCJ2022-34 |
発行日 |
2022-07-08 (EMCJ) |
ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
EMCJ2022-34 |