In the implementation of BCAB, why does the ``series_out'' return to its input one? It seems that the cross-attention does not play a role in series representation.

nzl5116190 / Basisformer

This is the pytorch implementation of Basisformer in the Neurips paper: [BasisFormer: Attention-based Time Series Forecasting with Learnable and Interpretable Basis]

84 stars 8 forks source link

In the implementation of BCAB, why does the ``series_out'' return to its input one? It seems that the cross-attention does not play a role in series representation. #5

Closed zzwzz1998 closed 6 months ago

zzwzz1998 commented 11 months ago

In the implementation of BCAB, why does the ``series_out'' return to its input one? It seems that the cross-attention does not play a role in series representation.

nzl5116190 commented 6 months ago

Thank you for pointing this out. Indeed, the line returning series_out to its input was meant to be commented out. The latest version of the code has addressed this issue, and we appreciate your vigilance in catching this oversight. Thanks again for your contribution to improving the project!