Summary of Improving Robustness to Corruptions with Multiplicative Weight Perturbations, by Trung Trinh et al.
Improving robustness to corruptions with multiplicative weight perturbations
by Trung Trinh, Markus Heinonen, Luigi Acerbi, Samuel Kaski
First submitted to arxiv on: 24 Jun 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper introduces an alternative approach called Data Augmentation via Multiplicative Perturbation (DAMP) that improves the robustness of deep neural networks (DNNs) to a wide range of corruptions without compromising accuracy on clean images. The authors demonstrate that input perturbations can be mimicked by multiplicative perturbations in the weight space and propose DAMP as a training method that optimizes DNNs under random multiplicative weight perturbations. The approach is evaluated on image classification datasets (CIFAR-10/100, TinyImageNet, and ImageNet) and neural network architectures (ResNet50, ViT-S/16, and ViT-B/16), showing enhancements in model generalization performance in the presence of corruptions across different settings. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about finding a way to make deep learning models more robust to noisy or corrupted images. Right now, these models are great at recognizing clean images but struggle when the image is messed up. The researchers found that by making tiny changes to the model’s internal weights, they can help it generalize better even when the input images are distorted. This approach, called DAMP, helps train models that perform well on both normal and corrupted images. |
Keywords
» Artificial intelligence » Data augmentation » Deep learning » Generalization » Image classification » Neural network » Vit
