We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? # to your account
根据您的代码复现出来,在Humidity数据集上的结果和论文中的结果差了一个数量级,您在论文中写的结果是MAE:0.4531,复现出来的结果是MAE:4.531左右。而且根据您在arxiv上的Figure3(c)中,也很明显的显示出MAE的横坐标是从1.0-5.0,那最终结果反而是0.4531。
The text was updated successfully, but these errors were encountered:
实验部分有单位介绍,为了将单位的小数位数统一,因此表格中的单位并非实际单位。
Sorry, something went wrong.
感谢作者解惑!我还有一个问题,本文是没有采用滑动窗口的形式吗,而是使用每天的前12步预测后12步(看generate data文件看不太懂)。
论文中采用的是2010-2018,链接提供的下载文件只有2010-2016的
一旦使用滑动,就会出现标签泄漏。即需要预测的输出会在下一次作为输入,被模型所见。
因此这里直接采用按24小时分割,避免标签泄漏
No branches or pull requests
根据您的代码复现出来,在Humidity数据集上的结果和论文中的结果差了一个数量级,您在论文中写的结果是MAE:0.4531,复现出来的结果是MAE:4.531左右。而且根据您在arxiv上的Figure3(c)中,也很明显的显示出MAE的横坐标是从1.0-5.0,那最终结果反而是0.4531。
The text was updated successfully, but these errors were encountered: