期刊
PROCEEDINGS OF THE GREAT LAKES SYMPOSIUM ON VLSI 2023, GLSVLSI 2023
卷 -, 期 -, 页码 483-488出版社
ASSOC COMPUTING MACHINERY
DOI: 10.1145/3583781.3590251
关键词
Microfluidic biochips; sample preparation; Graph attention network
Microfluidic biochips are promising and versatile in automating biochemical protocols. Machine learning models can enhance the efficiency and scalability of concentration prediction for microfluidic mixers, but struggle with multiple input flow rates. This paper proposes a new concentration prediction method based on the graph attention networks (GAT), which efficiently predicts the generated concentration of random microfluidic mixers with multiple input flow rates.
Microfluidic biochips have emerged with significant promise and versatility in automating a variety of biochemical protocols. Accurate preparation of fluid samples with microfluidic mixers is an essential component of these protocols, where concentration prediction and generation are critical. Recently, machine learning models have been adopted in concentration prediction, which demonstrate great potential in enhancing the efficiency and scalability over the traditional finite element analysis (FEA) methods. However, the state-of-the-art machine learning-based method can only predict the concentration of microfluidic mixers with fixed input flow rates, but suffers poor prediction accuracy for multiple input flow rates. To address this issue, this paper proposes a new concentration prediction method based on the graph attention networks (GAT). By modeling each channel of the mixer as a graph node in a GAT, the proposed method efficiently and accurately predicts the generated concentration of random microfluidic mixers with multiple input flow rates. Experimental results show that compared with the state-of-the-art method, the proposed GAT-based simulation method obtains a reduction of 85% in terms of errors of predicted concentration, which validates the effectiveness of the proposed GAT model.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据