.png){kind=icon}
返回
An Efficient Training Data Collection Method for Machine Learning-Based Frequency Selective Surface Design
.png){kind=icon}
PRE
AI
DOI：10.1109/LAWP.2024.3456838.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
To enhance the efficiency of the training dataset construction and improve the machine learning (ML) model performance for electromagnetic (EM) devices modeling and design, an efficient training data collection method based on the evolutionary algorithm is proposed. By setting an appropriate objective function for EM response, the evolutionary algorithm guides the training samples to contain more helpful and useful information for design in each optimization iteration. Consequently, with this higher-quality dataset, the ML model achieves better performance more readily. To fully demonstrate the validity of this proposed evolutionary algorithm-based training data collection method, a topological design example for frequency selective surface (FSS) with different incident angles is presented. Results indicate that, with the same number of samples, when compared with the traditional random data collection method, the proposed method improves testing accuracy by a maximum value of 29.6%. Furthermore, if the traditional random training data collection method is used to achieve the same testing error level as the proposed training data collection method, it would require more than twice the number of full-wave EM simulations.
Keyword:
Evolutionary algorithm
frequency selective surface (FSS)
machine learning (ML)
training data collection method
Evolutionary algorithm
frequency selective surface (FSS)
machine learning (ML)
training data collection method
期刊
IF:
4.8
论文数:
1.0W
被引数:
2.8W
机构
引用论文
.png){kind=icon}
Post‐stroke depression and functional recovery in a population‐based stroke register. The Finnstroke study基于人群的卒中登记中的卒中后抑郁和功能恢复。Finnstroke研究
PREAI
分享收藏Convolutional Neural Network for Behavioral Modeling and Predistortion of Wideband Power Amplifiers
OAAI
OAAI分享收藏
查看更多内容