GCformer combines a structured global convolutional branch for processing long input sequences with a local Transformer-based branch for capturing short, recent signals. Experiments demonstrate that GCformer outperforms state-of-the-art methods, reducing MSE error in multivariate time series benchmarks by 4.38% and model parameters by 61.92%. In particular, the global convolutional branch can serve as a plug-in block to enhance the performance of other models, with an average improvement of 31.93%, including various recently published Transformer-based models.
 |
|---|
| Figure 1. GCformer overall framework |
 |
|---|
| Figure 2. Different parameterization methods of global convolution kernel |
|
|
|
|
- Install Python 3.6, PyTorch 1.11.0.
- Download data. You can obtain all the six benchmarks from [FEDformer] or [Autoformer].
- Train the model. We provide the experiment scripts of all benchmarks under the folder
./scripts/GCformer. For instance, you can reproduce the experiment result on illness dataset by:
bash ./scripts/GCformer/illness.shWe appreciate the following github repos a lot for their valuable code base or datasets:
https://github.com/yuqinie98/PatchTST
https://github.com/MAZiqing/FEDformer