School of Data Science and Artificial Intelligence, Jilin Engineering Normal University, Changchun, China.
College of Information Technology, Jilin Agricultural University, Changchun, China.
PLoS One. 2024 Aug 16;19(8):e0305653. doi: 10.1371/journal.pone.0305653. eCollection 2024.
The setting of parameter values will directly affect the performance of the neural network, and the manual parameter tuning speed is slow, and it is difficult to find the optimal combination of parameters. Based on this, this paper applies the improved Hunger Games search algorithm to find the optimal value of neural network parameters adaptively, and proposes an ATHGS-GoogleNet model. Firstly, adaptive weights and chaos mapping were integrated into the hunger search algorithm to construct a new algorithm, ATHGS. Secondly, the improved ATHGS algorithm was used to optimize the parameters of GoogleNet to construct a new model, ATHGS-GoogleNet. Finally, in order to verify the effectiveness of the proposed algorithm ATHGS and the model ATHGS-GoogleNet, a comparative experiment was set up. Experimental results show that the proposed algorithm ATHGS shows the best optimization performance in the three engineering experimental designs, and the accuracy of the proposed model ATHGS-GoogleNet reaches 98.1%, the sensitivity reaches 100%, and the precision reaches 99.5%.
参数值的设定将直接影响神经网络的性能，而手动参数调整速度较慢，很难找到参数的最佳组合。基于此，本文将改进的 Hunger Games 搜索算法应用于自适应寻找神经网络参数的最优值，并提出了一种 ATHGS-GoogleNet 模型。首先，自适应权重和混沌映射被集成到饥饿搜索算法中，构建了一种新的算法 ATHGS。其次，改进的 ATHGS 算法被用于优化 GoogleNet 的参数，构建了一个新的模型 ATHGS-GoogleNet。最后，为了验证所提出的算法 ATHGS 和模型 ATHGS-GoogleNet 的有效性，设置了一个对比实验。实验结果表明，所提出的算法 ATHGS 在三个工程设计实验中表现出最佳的优化性能，而所提出的模型 ATHGS-GoogleNet 的准确率达到 98.1%，灵敏度达到 100%，精度达到 99.5%。
