[Re] What do neural networks learn in image classification? A frequency shortcut perspective

Purdue University ML 2023 Hackathon Reproducibility Challenge Submission1 Authors

Keywords: Computer Vision, Explainable AI

Abstract: The original paper of this replication presents an empirical study of the learning dynamics of neural networks for classification tasks through the use of frequency analysis. Our goal is to reproduce experimental results presented by the authors in the paper, which consist of frequency analysis conducted on both synthetic datasets and the ImageNet‐10 dataset.

Submission Number: 1