Even got better results? How?

[breadacker]

I had tried running FreTS model on ETTm1 and ETTh1 datasets with both univariate and multivariate forecasting tasks. The following results are even better than the results mentioned in your article. How is that possible?

---Hyperparameters---
Used MS or S for features
Used freq as detailed like 1min rathen than t as minutely
Used other settings as default given by you

---GPU--
RTX 3060 6GB

Univariate ETTh1 Test Results
mse:0.002602622378617525, mae:0.03822975605726242, rmse:0.05101590231060982

Multivariate ETTh1 Test Results
mse:0.002601700136438012, mae:0.038223933428525925, rmse:0.05100686475634575

Univariate ETTm1 Test Results
mse:0.0011898829834535718, mae:0.025772279128432274, rmse:0.03449467942118645

Multivariate ETTm1 Test Results
mse:0.0012134966673329473, mae:0.026144789531826973, rmse:0.03483527898788452

[abaklykov]

Where did you get these two datasets?

[breadacker]

https://github.com/zhouhaoyi/ETDataset/tree/main/ETT-small

Where did you get these two datasets?

[abaklykov]

Wow! Thanks