ryanchankh / zmzm-iclr-2020

Code for "Understanding l4-Based Dictionary Learning: Interpretation, Stability, and Robustness."

Understanding l4-based Dictionary Learning: Interpretation, Stability, and Robustness

Yuexiang Zhai, Hermish Mehta, Zhengyuan Zhou, Yi Ma

This repository contains code for the numerical and image experiments presented in the paper.

Description

Folder image-experiments contains codes and results for all image experiments in Figure 5, 6, and 7.

Folders msp-comparison, rtr-comparison, and sg-comparison contain codes for reproducing comparison experiments in Table 2.

The Riemannian subgradient descent and trust-region method implementations used are based on those provided by the original authors [1, 2]; the original code, along with our modifications, can be found in the appropriate folders. For any questions or bug reports, please feel free to reach out!

---

[1] https://github.com/sunju/ODL_L1

[2] https://github.com/sunju/dl_focm