A PSO-based fine-tuning algorithm for CNN
Tianyang Li, Haoyan Luo, Chenyu Wu · 2021 5th Asian Conference on Artificial Intelligence Technology (ACAIT) · 2021
With the continuous development of artificial intelligence, the Convolutional Neural Network (CNN) has been successfully applied to solve the classification problems. However, with a small sample data set that has large-scale of data features, CNN may have poor classification performances. With this concern, we focus on the classification problem with CNN on small sample data sets and a particle swarm optimization (PSO)-based algorithm for fine-tuning CNN is proposed. This algorithm considers the gradient direction to directly optimize weights of CNN, and can help CNN stepping out of stagnation effectively with less training data. Experimental results under a real world data set show the advantage of our proposal, in terms of minimized loss and maximized accuracy. The code is available at https://github.com/luohaoyan/CS-GA-W-PSO