Research on the Complexity Characteristics of Convolutional Neural Networks

Research on the Complexity Characteristics of Convolutional Neural Networks

연구 분야: Artificial Intelligence

학회: 2023 IEEE 7th Information Technology and Mechatronics Engineering Conference (ITOEC)

In order to quickly realize the convolutional neural network that meets the given complexity characteristics, a generation algorithm from complex network topology to convolutional neural networks is given. By establishing a series of convolutional neural networks with different topology characteristics, CIFAR 10 and CIFAR10 dataset analysis uses analysis The effects of topological properties such as average aggregation coefficient, average path length, diagram density, and modularity affect the effectiveness of the validity of convolutional neural network recognition. Experiments show that when the number of parameters of the neural network is basically equal, the average cluster coefficient will affect the performance of the convolutional neural network. In the end, in the statistically significant conclusion, the network structure with a small concentration coefficient will have better performance, which provides a theoretical basis for further designing a better convolutional neural network.