Meta-forests

• Description
• Installation
• Datasets
• Citation

This repository contains the code for this paper.

Setup a virtual environment with Python 3.10 and install CUDA with version 12.6+:

# sample command with conda virtual environment creation
conda create -n meta-forests python=3.10
conda activate meta-forests

Install the requirements:

pip install -r requirements.txt

Or install with command:

pip install deeplake==4.1.10 scikit-learn==1.6.1 numpy==2.2.4
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu126

⚠️ Please be reminded that Deeplake v4 does not support Windows, please proceed with WSL if you are on Windows.

We have used the following datasets according to the paper:

• VLCS
• PACS

While the blood glucose monitoring dataset from Sun et al. (2023) used in the paper is not publicly available.

1. Download the VLCS dataset from the link.
2. Decompress the file and place the VLCS folder in the src/data folder.
3. Download the AlexNet weights from the link and place the alexnet_caffe.pth.tar file in the src/weights folder.
4. Download the PACS dataset with the function load_pacs_training_dataset() in src/load_data.py.

@misc{sun2024metaforests,
      title={Meta-forests: Domain generalization on random forests with meta-learning}, 
      author={Yuyang Sun, Panagiotis Kosmas},
      year={2024},
      eprint={2401.04425},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2401.04425}, 
}

This project is licensed under the MIT License. See the LICENSE file for details.