Summary of Dataset Distribution Impacts Model Fairness: Single Vs. Multi-task Learning, by Ralf Raumanns et al.
Dataset Distribution Impacts Model Fairness: Single vs. Multi-Task Learning
by Ralf Raumanns, Gerard Schouten, Josien P. W. Pluim, Veronika Cheplygina
First submitted to arxiv on: 24 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 investigates how bias in datasets affects the fairness of model predictions in skin lesion classification using ResNet-based CNNs. The study focuses on patient sex variations in training data and evaluates three learning strategies: single-task, reinforcing multi-task, and adversarial learning. Results show that sex-specific training data yields better results, but single-task models exhibit sex bias. The reinforcement approach does not remove sex bias, while the adversarial model eliminates sex bias for female patients only. Datasets including male patients enhance model performance for the male subgroup even when females are the majority. Future research will explore more demographic attributes and confounding factors. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how biased datasets can affect the accuracy of computer models in diagnosing skin lesions. The study uses a type of AI called CNNs to test different ways of training the model, focusing on whether the model performs better when trained with data from one sex or both sexes. The results show that when the model is only trained with data from one sex, it does better and doesn’t have bias towards that sex. However, if the model is trained with data from both sexes, it can still be biased towards one sex. To make sure models are fair to everyone, we need to include more information about different demographics in our training data. |
Keywords
» Artificial intelligence » Classification » Multi task » Resnet
