Support Vector Machine Classifier Via 0-1 Loss Function

In the era of big data,whether it is multimedia,graphics,network communication,software engineering,and currently very hot artificial intelligence,machine learning technology can be found,especially in "computer application technology" fields such as computer vision and natural language processing.As a machine learning technology,support vector machine has been in a rapid development process,especially the soft margin support vector machine constructed after introducing the concepts of loss function and soft margin.It has always been the focus of scholars' research.Many loss functions with different structures have been proposed one after another.At the same time,some fast and effective algorithms have been designed,which provide a lot of help to solve the problem of data classification and are worthy of our in-depth discussion and careful study.Starting from the support vector machine model based on 0-1 loss,this paper first introduces the preparatory knowledge and the main work of this paper,then briefly introduces five commonly used soft margin support vector machine models,shows the expression and image of the corresponding loss function,analyzes the advantages and disadvantages of each model,summarizes the algorithms for solving different types of support vector machine models,and gives the corresponding iterative framework.On this basis,it makes a comparative study with L0/1-SVM,analyzes some theoretical properties of the model,and gives the sufficient conditions for the solution and the firstorder optimality conditions of the model.At the same time,a fast and stable algorithm,L0/1-ADMM,is designed to solve L0/1-SVM.The process of solving each subproblem in our algorithm and the skills used are explained in detail.Finally,we compare our model with other five soft margin support vector machines in classification accuracy and efficiency.A large number of numerical experiments prove that our model and classification algorithm achieve good numerical results in accuracy and computational efficiency,both on artificial data sets and real data sets.