Summary of Effective and Robust Adversarial Training Against Data and Label Corruptions, by Peng-fei Zhang et al.
Effective and Robust Adversarial Training against Data and Label Corruptions
by Peng-Fei Zhang, Zi Huang, Xin-Shun Xu, Guangdong Bai
First submitted to arxiv on: 7 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
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 proposes an Effective and Robust Adversarial Training (ERAT) framework that can handle both data corruptions and label noise in datasets from unreliable sources. Existing methods often overlook the co-existence of these corruptions, limiting model effectiveness. The ERAT framework uses hybrid adversarial training with multiple potential perturbations to learn resilient models for dual corruption. This is achieved by generating surrogate malicious data perturbations using a deep neural network (DNN) model and maintaining semantic consistency between original and perturbed data. Additionally, a class-rebalancing data selection strategy discards noisy labels through semi-supervised learning. Experimental results demonstrate the ERAT framework’s superiority. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making artificial intelligence models more reliable when they’re trained on bad or noisy data. Imagine you have a computer program that can recognize pictures, but it was trained using some tricky images. You want the program to be able to handle those tricky images and still work well with normal pictures. The authors of this paper created a new way to train these models called ERAT (Effective and Robust Adversarial Training). It’s like having a special kind of armor for your model that helps it stay strong even when faced with bad data. |
Keywords
» Artificial intelligence » Neural network » Semi supervised
