Summary of Disparate Impact on Group Accuracy Of Linearization For Private Inference, by Saswat Das et al.
Disparate Impact on Group Accuracy of Linearization for Private Inference
by Saswat Das, Marco Romanelli, Ferdinando Fioretto
First submitted to arxiv on: 6 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR); Computers and Society (cs.CY)
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 explores the trade-off between computational efficiency and fairness in neural networks that rely on cryptographically secure data. By linearizing certain activations to reduce runtime, researchers have found significant performance gains with minimal accuracy loss. However, this approach may inadvertently exacerbate fairness issues by disproportionately affecting minority groups’ accuracy compared to majority groups. The authors provide a mathematical explanation for these findings and demonstrate their prevalence across various datasets and architectures. They also propose a simple mitigation strategy that adjusts the fine-tuning process for linearized models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A team of researchers investigated how making neural networks more efficient can affect fairness. They found that when they reduced the complexity of certain parts of the network to speed up processing, it actually made things worse for minority groups. This is because these changes can make it harder for the network to accurately identify minority group members. The authors show that this problem occurs in many real-world datasets and models. To fix this issue, they suggest a simple tweak to the way the model is fine-tuned. |
Keywords
* Artificial intelligence * Fine tuning
