Critical Periods Emerge Even in Deep Linear Networks

Official Pytorch Implementation for Critical Periods Emerge Even in Deep Linear Networks (ICLR 2024) .

Quick Start

Our code is organized into two repositiories: multipath and matrixcompletion. Each repository contains a README.md file with instructions on how to reproduce the results in the paper.

Requirements

• python
• torch
• numpy
• scipy
• matplotlib
• pyyaml
• seaborn

Acknowledgements

This repository builds off:

• https://github.com/roosephu/deep_matrix_factorization
• https://github.com/AllenInstitute/Multipathway_NeurIPS2022

---

If you find this useful for your work, please consider citing

@inproceedings{
kleinman2024critical,
title={Critical Learning Periods Emerge Even in Deep Linear Networks},
author={Michael Kleinman and Alessandro Achille and Stefano Soatto},
booktitle={The Twelfth International Conference on Learning Representations},
year={2024},
url={https://openreview.net/forum?id=Aq35gl2c1k}
}