This repository contains code and scripts to recreate the task-similarity, quadratic growth, and deep learning experiments from the FMRL paper (citation below).
- task-similarity: similarity.py
- quadratic growth: growth.py
- deep learning: .sh scripts in reptile directory (edited clone of OpenAI's Reptile codebase)
- MiniWiki dataset: .tar.gz file in data/
- MiniWiki experiments: online.py (Dependencies: scikit-learn, cvxpy, scipy, numpy, torch, h5py, matplotlib, text_embedding)
Citation:
@inproceedings{khodak2019fmrl,
title={Provable Guarantees for Gradient-Based Meta-Learning},
author={Khodak, Mikhail and Balcan, Maria-Florina and Talwalkar, Ameet},
booktitle={Proceedings of the 36th International Conference on Machine Learning,},
year={2019}
}