Summary of Robust Black-box Testing Of Deep Neural Networks Using Co-domain Coverage, by Aishwarya Gupta et al.
Robust Black-box Testing of Deep Neural Networks using Co-Domain Coverage
by Aishwarya Gupta, Indranil Saha, Piyush Rai
First submitted to arxiv on: 13 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 A novel black-box approach is proposed to generate test-suites for robust testing of deep neural networks (DNNs). The method, called Co-Domain Coverage (CDC), defines a coverage criterion that takes into account the end-to-end behavior of the model’s output. A new fuzz testing procedure, CoDoFuzz, uses CDC to guide the generation of test inputs and is compared with state-of-the-art methods for DNNs trained on six publicly available datasets. Experimental results demonstrate the efficiency and efficacy of CoDoFuzz in generating a large number of misclassified inputs and inputs where the model lacks confidence. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A machine learning paper is looking at how to make sure deep neural networks (DNNs) are reliable. The researchers developed a new way to test DNNs without knowing how they work inside. This approach, called Co-Domain Coverage, looks at what the model produces and uses that to create tests. They then created a new testing tool, CoDoFuzz, which uses this method to find problems with the model. They tested their approach on six different datasets and showed that it works well. |
Keywords
» Artificial intelligence » Machine learning
