Applied Research on Face Recognition Based on Transformer and Deep Reinforcement Learning

The progress of deep learning technology has greatly enhanced the performance of face recognition system, but there are also limitations such as low sample efficiency, unstable training, etc. Transformer is a revolutionary deep learning model architecture with the advantages of parallel computing and powerful long-range dependent modelling capability. This paper analyses the advantages and disadvantages of classical methods, deep reinforcement learning, Transformer and other methods applied in face recognition after reviewing many literatures, and concludes that the application of deep reinforcement learning combined with Transformer in face recognition is still in the early stage, but it has also achieved certain results, and it is a direction full of development potential in the field of artificial intelligence today.