Performance Analysis of Fine-tune Transferred Deep Learning

Abstract

Nowadays, artificial intelligence (AI) and deep learning becomes the most important research issues. The history of AI technology started from machine learning, convolution neural networks (CNN), recurrent neural network (RNN), long short-term memory (LSTM), then developed to the latest technology of generative adversarial network (GAN). Deep learning is improving quickly. Data augmentation and transfer learning are currently advanced functions in the field of deep learning, which accelerates the whole training of the neural networks and improves model accuracy. This study implements these two functions and analyzes the performance in a deep learning model. Due to our performance study, researchers can make the most efficient AI model. To find a better deep learning model, we evaluate five deep learning strategies. In this way, we compare different learning effects and observe model accuracy. According to analysis charts and data tables, this study shows that transferred deep learning and fine-tune strategy not only increase the accuracy but also give insignificant overfitting. We also prove that fine-tune transfer learning significantly saves more time and costs in future tests and experiments.