liguge / AAU_Net_Improved

The improved code of AAU-Net

AAU-Net

1. We provide the improved code for paper "Adversarial Algorithm Unrolling Network for Interpretable Mechanical Anomaly Detection".

2. Compared with the original Code[https://github.com/Botao-An/AAU_Net], Zhi Cao[[e-mail]] improved the loss functions of AAU-Net and solved the problem of instability during training. The referred paper is "Wasserstein Auto-Encoders"

3. The code is originally debugged on the computer with the following configuration.

   Hardware	Software	Software
   Intel Core i7-10700KF CUP	Anaconda 4.9.2	CUDA 11.7
   RAM 32GB	Python 3.8	cuDNN 8.4.0
   NVIDIA GeForce RTX 3080 GPU	PyTorch 1.8.1	TorchVision 0.9.1

4. We give a demo of simulation for you to test the performance of AAU-Net initially.

5. To run the model, you should firstly prepare the dataset in a standard format as the simulated one. More details can be found in file pre and post processing.ipynb. Then you can run AAU-Net by file train.py. To check and visulize the results, you can use pre and post processing.ipynb again.

6. If you want do research based this code, please cite the following papers. For any questions you can contact e-mail: [e-mail] or [e-mail].

   @article{an2023interpretable,
      title={Interpretable Neural Network via Algorithm Unrolling for Mechanical Fault Diagnosis},
      author={An, Botao and Wang, Shibin and Zhao, Zhibin and Qin, Fuhua and Yan, Ruqiang and Chen, Xuefeng},
      journal={IEEE Transactions on Instrumentation and Measurement},
      volume={71},
      pages={1--11},
      year={2023},
      publisher={IEEE}
   }
   @article{an2022adversarial,
      title={Adversarial Algorithm Unrolling Network for Interpretable Mechanical Anomaly Detection},
      author={An, Botao and Wang, Shibin and Qin, Fuhua and Zhao, Zhibin and Yan, Ruqiang and Chen, Xuefeng},
      journal={IEEE Transactions on Neural Networks and Learning Systems},
      volume={},
      pages={},
      year={2022},
      publisher={IEEE}
   }