![[2401.07390] Knee or ROC](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2401.07390] Knee or ROC
Summary
The paper 'Knee or ROC' explores accuracy measurement methods for multi-class image detection using self-attention transformers, proposing a knee threshold approach as an alternative to traditional ROC methods.
Why It Matters
This research addresses limitations in current accuracy testing for multi-class image detection, particularly when the representation of image populations is unknown. By proposing the knee method, it offers a potentially more adaptable solution for practitioners in machine learning and computer vision.
Key Takeaways
- Self-attention transformers show promise in image classification with limited data.
- Traditional ROC methods may not be suitable for multi-class input images with unknown population representation.
- The knee method provides a flexible alternative for determining accuracy thresholds.
Computer Science > Machine Learning arXiv:2401.07390 (cs) [Submitted on 14 Jan 2024 (v1), last revised 23 Feb 2026 (this version, v2)] Title:Knee or ROC Authors:Veronica Wendt, Jacob Steiner, Byunggu Yu, Caleb Kelly, Justin Kim View a PDF of the paper titled Knee or ROC, by Veronica Wendt and 4 other authors View PDF HTML (experimental) Abstract:Self-attention transformers have demonstrated accuracy for image classification with smaller data sets. However, a limitation is that tests to-date are based upon single class image detection with known representation of image populations. For instances where the input image classes may be greater than one and test sets that lack full information on representation of image populations, accuracy calculations must adapt. The Receiver Operating Characteristic (ROC) accuracy threshold can address the instances of multiclass input images. However, this approach is unsuitable in instances where image population representation is unknown. We then consider calculating accuracy using the knee method to determine threshold values on an ad-hoc basis. Results of ROC curve and knee thresholds for a multi-class data set, created from CIFAR-10 images, are discussed for multiclass image detection. Comments: Subjects: Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2401.07390 [cs.LG] (or arXiv:2401.07390v2 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2401.07390 Focus to learn more arXiv-issue...
