Confusion about SMD dataset

[1220812196]

I am honored to read your paper: GDformer: Going Beyond Subsequence Isolation for Multivariate Time Series Anomaly Detection, As for the results of SMD data set, through my own adjustment of your code, the result is about 84.7, which is far from the experimental result of 90.41 in your paper. I would like to ask how to reproduce the results close to yours? Because I feel that there is still a big gap between the results I reproduced and the actual results of your paper. Please help me solve my doubts.

[Peachypie98]

I haven't check the code throughly, it seems like the author used 1D CNN instead of linear layers for input embeddings. Also, the mask ratio is not implemented (not being used atm).

I haven't tried reproducing the results using SMD dataset, but I tried using MSL dataset and still can't get the same results in the paper

(update - 250702) they did implement the mask within Solver.train() function