Summary of Dirichlet-based Prediction Calibration For Learning with Noisy Labels, by Chen-chen Zong et al.
Dirichlet-Based Prediction Calibration for Learning with Noisy Labels
by Chen-Chen Zong, Ye-Wen Wang, Ming-Kun Xie, Sheng-Jun Huang
First submitted to arxiv on: 13 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 proposed Dirichlet-based Prediction Calibration (DPC) method addresses the issue of noisy labels in deep neural networks (DNNs). Existing approaches rely on the softmax function, which can be over-confident and unreliable. The DPC method introduces a calibrated softmax function that breaks translation invariance by incorporating a suitable constant in the exponent term. This enables more reliable model predictions. To ensure stable model training, the proposed evidence deep learning (EDL) loss is introduced. This loss encourages positive and large logits for the given label while penalizing negative and small logits for other labels. Extensive experiments on benchmark datasets demonstrate state-of-the-art performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A team of researchers found a way to improve how artificial intelligence models learn from noisy data. They noticed that existing methods were not reliable because they relied on a specific math function called softmax. The new approach, called Dirichlet-based Prediction Calibration (DPC), uses a different kind of math function to make AI models more accurate. This helps the models avoid mistakes caused by noisy data. The researchers tested their method with many examples and showed that it performs better than other methods. |
Keywords
* Artificial intelligence * Deep learning * Logits * Softmax * Translation
