Summary of Dual-label Learning with Irregularly Present Labels, by Mingqian Li et al.
Dual-Label Learning With Irregularly Present Labels
by Mingqian Li, Qiao Han, Yiteng Zhai, Ruifeng Li, Yao Yang, Hongyang Chen
First submitted to arxiv on: 18 Oct 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 This paper introduces Dual-Label Learning (DLL), a novel framework for two-label learning that effectively utilizes irregularly present labels in multi-task learning. The DLL framework formulates the problem as a dual-function system, incorporating standard supervision, structural duality, and probabilistic duality to capture label correlation. A dual-tower model architecture is designed to explicitly exchange information between labels, allowing for label imputation during training and joint solution of unknown labels during inference. Theoretical analysis confirms the feasibility of DLL, and experimental results demonstrate a significant improvement in F1-score or mean absolute percentage error (MAPE) compared to baseline approaches, with up to 10% gains. Interestingly, DLL can achieve similar or better results than baselines even at high label missing rates (up to 60%). |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about how we learn from data when some of the labels are missing. In drug analysis, for example, we might know some properties of a drug but not others because it’s hard to get all that information. The researchers propose a new way to do this called Dual-Label Learning (DLL). It’s like solving two puzzles at once: one for each type of label. They show that by doing this, they can make better predictions than usual, even when lots of labels are missing. |
Keywords
» Artificial intelligence » F1 score » Inference » Multi task
