摘要
为了解决无源光纤环形谐振器动力学仿真中传统数值方法计算效率低的问题,采用一种基于动态投票机制的混合机器学习模型,通过融合梯度提升决策树(GBDT)、支持向量机(SVM)和人工神经网络(ANN)实现谐振器内光场状态的快速准确分类。基于训练集分别训练GBDT、SVM和ANN模型,并通过动态加权投票,以多数表决或基于类别准确率的权重择优机制确定最终输出。结果表明,模型在噪声态、图灵态、混沌态、单孤子态和多孤子态的分类准确率分别达到77%、95%、97%、86%和85%;相比随机森林,后两类准确率提升超过12%,且计算效率较传统Lugiato-Lefever方程提升万倍以上。该方法不仅解决了传统数值方法的效率问题,还通过多模型协同优化实现了相比单一模型更稳健的分类性能,为智能算法替代复杂数值仿真、推动非线性光子学智能化发展提供了新思路。
The dynamical simulation of passive fiber ring resonators typically relies on the traditional Lugiato-Lefever equation(LLE)numerical method,which incurs high computational costs due to its iterative nature.This inefficiency poses a critical bottleneck for predicting optical field states within resonators.To address this challenge,this study proposes a hybrid machine learning model that integrates gradient boosting decision trees(GBDT),support vector machines(SVM),and artificial neural networks(ANN)through a dynamic voting mechanism.The primary objective is to achieve rapid and accurate classification of optical field states in resonators,thereby replacing traditional numerical simulations and advancing intelligent development in nonlinear photonics.The training dataset was generated by numerically solving the LLE under varying parameters,including fiber length(10 m~100 m),power transmission coefficient(0.05~0.15),and detuning(–π~π).The dataset encompassed five optical field states:noise,Turing,chaotic,single soliton,and multiple soliton states(Fig.2).Three machine learning models—GBDT,SVM,and ANN—were independently trained on this dataset.A dynamic weighted voting strategy was applied to synthesize their predictions:the final output was determined either by majority voting or by selecting the optimal result from the model with the highest historical accuracy for a specific class,combined with a weighting mechanism based on class-specific accuracy.Hyperparameter optimization using the Optuna framework ensured optimal performance,with key parameters including GBDT’s learning rate(0.01)and maximum tree depth(5),SVM’s penalty factor(C=90),radial basis function kernel,and kernel parameter(=35),and ANN’s architecture(three hidden layers with 60 neurons each).Results demonstrated classification accuracies of 77%(noise state),95%(Turing state),97%(chaotic state),86%(single soliton state),and 85%(multiple soliton states)for the hybrid model(Fig.6).Compared to random forest(RF),the hybrid model improved accuracy by over 12%for single-and multiple soliton states.Confusion matrices(Fig.5)revealed that GBDT excelled in single-soliton classification,SVM outperformed in multiple soliton identification,and ANN achieved superior noise-state recognition.The dynamic voting mechanism effectively balanced model strengths and mitigated individual weaknesses,reducing misclassification and enhancing robustness.Parameter space testing(Fig.7)confirmed strong consistency between predicted and actual state distributions.In computational efficiency tests(Fig.8),the hybrid model processed 100 parameter sets in 0.24 s—over 10000 times faster than the traditional LLE method(2700 s).By combining GBDT,SVM,and ANN with dynamic voting,this hybrid machine learning framework successfully resolves the efficiency limitations of LLE-based methods and improves upon single-model constraints,achieving unprecedented computational speed.The study highlights the potential of intelligent algorithms to replace complex numerical simulations in nonlinear photonics.
作者
顾琰恺
臧裕斌
朱润徽
王有为
周瑞
张祖兴
GU Yankai;ZANG Yubin;ZHU Runhui;WANG Youwei;ZHOU Rui;ZHANG Zuxing(College of Electronic and Optical Engineering,College of Flexible Electronics(Future Technology),Nanjing University of Posts and Telecommunications,Nanjing 210023,China)
出处
《激光技术》
北大核心
2026年第1期1-8,共8页
Laser Technology
基金
国家自然科学基金资助项目(62175116)。
关键词
光纤光学
无源光纤谐振器
机器学习
人工神经网络
分类识别
fiber optics
passive fiber optic resonator
machine learning
artificial neural network
classification and identification
