大家好，我初步对比了一下我们实验代码与TSLib之间的差别，主要变化在learning rate strategy，我们在整理算法库的时候额外增加了learning rate decay的过程，但是这一设计会降低模型训练的波动性，对于一些数据量较小的数据集反而不友好，因此我们在最近的commit中已经去掉了这一过程（1c7f843 ），我这边测试是可以复现结果的，请大家再次尝试
As mentioned in the issue, I add the following two lines from commit:1c7f843.
However, I am still not able to reproduce the result. Moreover, the average result dropped after making the amendment mentioned in the commit.
Here are my results using TimesNet:
|
Table 17 |
Reproduce |
commit #1c7f843 |
| EthanolConcentration |
35.7 |
28.9 |
28.1 |
| FaceDetection |
68.6 |
66.3 |
68.0 |
| Handwriting |
32.1 |
31.8 |
17.4 |
| Heartbeat |
78.0 |
77.1 |
76.1 |
| JapaneseVowels |
98.4 |
97.3 |
93.0 |
| PEMS-SF |
89.6 |
86.7 |
75.7 |
| SelfRegulationSCP1 |
91.8 |
89.8 |
90.1 |
| SelfRegulationSCP2 |
57.2 |
51.1 |
52.2 |
| SpokenArabicDigits |
99.0 |
99.2 |
98.8 |
| UWaveGestureLibrary |
85.3 |
88.1 |
85.6 |
| Avg |
73.6 |
71.6 |
68.5 |
Thank you for your help.
As mentioned in the issue, I add the following two lines from commit:1c7f843.
However, I am still not able to reproduce the result. Moreover, the average result dropped after making the amendment mentioned in the commit.
Here are my results using TimesNet:
Thank you for your help.