用专业学术英文优化：Various methods for with noisy labels have been proposed recently Among these methods the transition matrix method has attracted sustained attention due to its simplicity and statistical con

Recently, several methods have been proposed to deal with noisy labels. Among these methods, the transition matrix method has gained considerable attention due to its simplicity and statistical consistency. However, this method is mainly suitable for class-dependent label noise problems. In real-world scenarios with instance-dependent noise, estimating the transition matrix for each sample becomes challenging and computationally expensive using existing methods.

In this paper, we propose a simplified transition matrix method that only requires estimating a global matrix. This makes our approach applicable to various types of label noise, including instance-dependent noise. By estimating a global transition matrix, we can determine the overall probability transfer from correct labels to noisy labels. This estimation also helps in implicitly regularizing the difference between the posterior probability distribution and the noisy label distribution. Our approach can handle different types of noise and alleviate the problem of inaccurate posterior probability estimation.

We provide theoretical evidence to demonstrate the consistency and effectiveness of our proposed method. Additionally, we conduct experiments on synthetic and real datasets with various types of label noise. The experimental results clearly indicate that our method outperforms previous transition matrix methods and has a wider range of applicability. Moreover, our method achieves competitive results without the need for additional auxiliary techniques, as compared to other state-of-the-art methods. The code for our method is available at...