RVM is a kernel machine learning method based on Bayesian inference learning. It has the same sparse solution as SVM, so it only depends on the kernel function calculated on a subset of the training data when making predictions on new data. This subset is called the relevance vector (relevance vector). By introducing the Bayesian method, RVM can provide the posterior probability output of the test sample and can generate a more sparse solution. The classification performance of RVM is similar to or even better than SVM, and the sparsity is usually much better than SVM, especially when the training sample set is large, the advantage is more obvious, so RVM is more suitable than SVM for solving online problems under large training sample sets. classification problem.