翻译成专业学术英文：近期各种不同的关于标签噪音学习的方法被提出在这些方法中与样本筛选和一些通过增强对比的方式进行不同转移矩阵作为一种简便并且具有统计一致性的方法得到长期关注。通常转移矩阵方法更多针对类别依赖噪音问题在样本依赖噪音情形下需要对每个样本的转移矩阵进行估计其可识别性难以保证并且在已有的一些方法中计算消耗很大。在本文中提出了一种关于转移矩阵的简便方法只需要近似估计一个整体的矩阵即可适用于包

Various methods for label noise learning have been proposed recently. Among these methods, the transfer matrix has attracted significant attention due to its simplicity and statistical consistency. Typically, the transfer matrix method is more suitable for class-dependent noise problems. In the case of sample-dependent noise, estimating the transfer matrix for each sample is challenging and computationally expensive in existing methods. In this paper, we propose a simplified approach for the transfer matrix that only requires the approximate estimation of a global matrix, making it applicable to various types of noisy labels, including sample-dependent noise. Specifically, by estimating a transfer matrix, we obtain the overall probability change of noise and assist with implicit regularization to adjust the difference between the posterior probability distribution and the noise label distribution. This approach can be applied to various types of noise and alleviates the issue of inaccurate posterior probability estimation. We theoretically demonstrate the consistency and effectiveness of this method and conduct experiments on synthetic and real datasets with different types of noise. The experimental results show that our method significantly improves upon previous transfer matrix methods and has a wider range of applicability. Furthermore, our method achieves competitive results without the need for additional auxiliary techniques. Our code is available at...