PyTorch implementation for the paper "Disentangling Multi-view Representations via Curriculum Learning with Learnable Prior" (IJCAI 2025)

Requirements

• python == 3.10.15
• torch == 2.1.0
• torchvision == 0.16.0
• scikit-learn == 1.5.2
• scipy == 1.14.1

We also export our conda virtual environment as CL2P.yaml. You can use the following command to create the environment.

conda env create -f CL2P.yaml

Dataset and model

You could find the Office-31 dataset we used in the paper from Baidu Netdisk, and the pre-trained models from Baidu Netdisk.

Usage

Training

To train the model, use the following command:

python train.py -f configs/Edge-MNIST.yaml

This will start the training process using the configuration specified in configs/Edge-MNIST.yaml.

Testing

To test the trained model, use the following command:

python test.py -f configs/Edge-MNIST.yaml

This will load the trained model and test it using the configuration specified in configs/Edge-MNIST.yaml.

Citation

If you find CL2P useful in your research, please consider citing:

@inproceedings{guo2025cl2p,
  title={Disentangling Multi-view Representations via Curriculum Learning with Learnable Prior},
  author={Guo, Kai and Wang, Jiedong and Peng, Xi and Hu, Peng and Wang, Hao},
  journal={Proceedings of the 34th International Joint Conference on Artificial Intelligence},
  year={2025}
}

Acknowledgements

The codes are based on MRDD. Thanks to the authors for their codes!