View in EDS

Classification-regression backpropagation neural network for efficient planar lightwave circuit design.

Saved in:

Bibliographic Details
Title:	Classification-regression backpropagation neural network for efficient planar lightwave circuit design.
Authors:	Yang, Chentong¹ (AUTHOR), Zheng, Yu¹ (AUTHOR) zhengyu@csu.edu.cn, Fan, Keyi¹ (AUTHOR), Liu, Shu¹ (AUTHOR), Zeng, Ke¹ (AUTHOR), Ouyang, Xinyu¹ (AUTHOR)
Source:	Optics Communications. Oct2026, Vol. 616, pN.PAG-N.PAG. 1p.
Subjects:	Back propagation, Multi-objective optimization, Machine learning, Integrated optics
Abstract:	To address the challenges of high-dimensional parameter coupling, multi-objective co-optimization, and the inadequate capability of existing approaches in planar lightwave circuit (PLC) design, this paper proposes a two-stage classification-regression backpropagation neural network (BPNN) optimization scheme. With a typical 1 × 3 PLC splitter chip as the design case, a dataset comprising 350 parameter-performance samples is constructed, and two lightweight fully connected networks are independently trained for classification and regression tasks. Results reveal that the classification accuracy exceeds 90%, and the regression absolute error remains below 0.01 dB. The ultimately optimized device achieves performance across the 1.27-1.65 μm wavelength band with an excess loss (EL) below 0.0653 dB, a wavelength-dependent loss (WDL) less than 0.0609 dB, and a uniformity loss (UL) under 0.0689 dB, thereby providing an efficient and viable new solution for the intelligent modeling and optimization of PLC devices. [ABSTRACT FROM AUTHOR]
	Copyright of Optics Communications is the property of Elsevier B.V. and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	To address the challenges of high-dimensional parameter coupling, multi-objective co-optimization, and the inadequate capability of existing approaches in planar lightwave circuit (PLC) design, this paper proposes a two-stage classification-regression backpropagation neural network (BPNN) optimization scheme. With a typical 1 × 3 PLC splitter chip as the design case, a dataset comprising 350 parameter-performance samples is constructed, and two lightweight fully connected networks are independently trained for classification and regression tasks. Results reveal that the classification accuracy exceeds 90%, and the regression absolute error remains below 0.01 dB. The ultimately optimized device achieves performance across the 1.27-1.65 μm wavelength band with an excess loss (EL) below 0.0653 dB, a wavelength-dependent loss (WDL) less than 0.0609 dB, and a uniformity loss (UL) under 0.0689 dB, thereby providing an efficient and viable new solution for the intelligent modeling and optimization of PLC devices. [ABSTRACT FROM AUTHOR]
ISSN:	00304018
DOI:	10.1016/j.optcom.2026.133356