摘要
卷积神经网络(CNN)自身结构的超参数对于分类问题中的准确率与效率有较大的影响,针对现有超参数优化方法多依赖传统组合,优化结果不彻底,导致模型分类效果不佳的状况,提出一种基于混合优化算法的CNN超参数优化方法。该方法根据CNN架构的结构特点选取超参数,然后采用粒子群优化算法(PSO)-梯度下降(GD)混合算法进行优化。在测试数据集上的实验结果表明:该方法在分类问题上具有较好的性能,提升了超参数的优化效率,避免了传统PSO算法易陷入局部最优的缺点。
The hyperparameters of CNN’s own structure have a great impact on the accuracy and efficiency of classification problems.Considering the fact that the existing hyperparameter optimization methods rely on traditional combinations and the optimization results are incomplete,which leads to poor classification effect of the model,a CNN hyperparameter optimization method based on hybrid optimization algorithm was proposed.In this method,the hyperparameters were selected according to the structural characteristics of the CNN architecture,and then the PSO-GD(gradient descent)hybrid algorithm was employed for the optimization.The experimental results on the test data set show that,the proposed method has better performance in the classification,improves the efficiency of hyperparameter optimization and avoids the shortcomings of traditional PSO algorithm that is easy to fall into local optimum.
作者
丁彧洋
DING Yu-yang(School of Internet of Things Engineering,Jiangnan University)
出处
《化工自动化及仪表》
CAS
2023年第6期875-882,共8页
Control and Instruments in Chemical Industry
关键词
PSO-GD混合算法
超参数优化
CNN
分类性能
优化效率
局部最优
PSO-GD hybrid algorithm
hyperparameter optimization
CNN
classification performance
optimization efficiency
local optimum
