Facial Expression Recognition Based On Transfer Learning Algorithm

With the advent of big data age,people can more easily get a lot of data.In addition,with the continuous development of machine learning,how to make computers have the ability of drawing inferences about other cases from one instance,and how to make big data play a better role have become very practical and valuable.In order to solve these problems,transfer learning has been raised and gotten more and more attention from people.There is an important assumption in conventional machine learning that the data required for training and the data required for the target must have the same distribution or from the same feature space.In real life,this assumption is difficult to achieve.Specifically,for a classification problem,if the sample of the training set and the target set do not have the same distribution,it can be understood that the source domain does not have the same feature space as the target domain.The traditional solution is to get more data to simulate the distribution of the target domain.But this method is so expensive.Transfer learning is an effective way to shorten the distance between two domains,so that we can approach the assumptions in traditional machine learning and train the model for the target data through source data.This paper makes a detailed study and summary of the transfer learning algorithm.Specific work is as follows,(1)This paper summarizes and studies the existing transfer learning methods,and the properties of the methods are compared.The fields of the various algorithms are also described in detail.(2)The commonly used algorithms such as transfer component analysis algorithm,geodesic flow kernel algorithm,subspace alignment algorithm,maximum independent domain adaptation algorithm and Information-theoretical learning algorithm are studied in detail,and the transfer learning method based on deep learning is also studied.(3)The mentioned transfer learning methods in the article are applied to the facial expression recognition,and use different databases to do the experiment.The transfer learning effectively solves the classification problem when the source domain and the target domain do not have the same feature space in facial expression recognition.(4)At the end of this paper,the problems in transfer learning are analyzed,and the future development is forecasted.