Hello, Dr. Liu, thank you for this article that has inspired me a lot. I have a little doubt: Considering that there is no ground truth data, so you artificially inject exceptions into the test set, but there is a problem that the original test data will contain some abnormal data, and this part of the exception should not be ignored. However, the original test set data is directly regarded as normal data in the paper and code. How do you think about this problem? thank you~
刘博士您好，感谢这篇文章给我带来很大启发。我有一点疑惑：考虑到没有ground truth数据，所以人工向测试集注入异常，但有个问题是原有测试集数据本就会包含一部分的异常数据，这部分异常不应该被忽视。但论文和代码中直接将原有测试集数据认为是正常数据了，请问您怎么看待这个问题的？谢谢~
Hello, Dr. Liu, thank you for this article that has inspired me a lot. I have a little doubt: Considering that there is no ground truth data, so you artificially inject exceptions into the test set, but there is a problem that the original test data will contain some abnormal data, and this part of the exception should not be ignored. However, the original test set data is directly regarded as normal data in the paper and code. How do you think about this problem? thank you~