Wenjie Hu

仅供非商业用途，MIT协议

we don't get the approvement to release the real-world dataset

数据是按不同时间粒度聚合的，没有做具体的特征工程，采用原数据输入，做了基本的归一化处理 label的含义paper原文中有描述

这是 输入的张量shape 不对，检查下数据输入

有一个无监督版本，主要修改的输出层，对图特征进行异常检测

重新update了project，可以再看一下。之前是老版本的模型，没有及时更新

This work focuses on the time series classification (TSC) task.

We test the model based on the continous data, at least for now. I think it can be extended to the mixture scenarios, if you can find some methods to...

his_len is the length of the inputed segments, orderd by time

Please show me the error output