MetaLearning-Lab

The code and methods offered in Awesome-META+: https://wangjingyao07.github.io/Awesome-Meta-Learning-Platform/

Methods

• Datasets: Data and Tasks related codes. For download links of all data (including benchmarks), see https://wangjingyao07.github.io/Awesome-Meta-Learning-Platform/5%20Datasets/Datasets/

• Meta Modules: Framework modules for meta-learning, for auxiliary viewing. Not described in the platform. At the same time, the project is not complete and will be supplemented in the future.

• Methods: Rapid application and deployment of meta-learning frameworks. All dataset settings, multitasking settings, and distributed training are integrated in this folder. For details on how to use it: https://wangjingyao07.github.io/Awesome-Meta-Learning-Platform/2%20Documentation/1%20maml_tutorial/

• Optimizer: Meta-learning framework optimization ideas

• Paper: "Methods" framework related paper

• Tests: test code

Important Notice

The Datasets (code) and Meta Modules (CODE) of this project need to be supplemented and improved, and will be maintained and iterated in the future. In order to facilitate downloading, the final version will be uploaded in the form of Baidu network disk and Google cloud disk.

The .rar and .zip files are extracted and packaged directly from the Linux server, and can be decompressed directly in the folder where the files are located.

!!! Any question or suggestion please let me know, I am honored to accept everyone's suggestions and opinions, Best wishes.

Cite

please cite when you use our platform (This article is only the first draft of the competition, as a reference, which will be updated later.)

@misc{wang2023awesomemeta,
      title={Awesome-META+: Meta-Learning Research and Learning Platform}, 
      author={Jingyao Wang and Chuyuan Zhang and Ye Ding and Yuxuan Yang},
      year={2023},
      eprint={2304.12921},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

Thanks!