Combining Graph Transformers Based Multi-Label Active Learning and Informative Data Augmentation for Chest Xray Classification

Multi-label learning is an essential component of supervised learning that aims to predict a list of relevant labels for a given data point. In the era of big data, characterized by the continuous generation of complex datasets, multi-label learning tasks, such as multi-label classification (MLC) and multi-label ranking, present significant challenges, capturing considerable attention across various domains. Some of the inherent challenges include high-dimensional features and labels, label dependency, and the existence of partial or missing labels, all of which render traditional methods ineffective. In recent times, there has been a notable surge in the adoption of deep learning (DL) techniques to address these challenges more adeptly in MLC. Notably, there is a burgeoning effort to harness the robust learning capabilities of DL for improved modelling of label dependencies and other inherent complexities in MLC. However, it is noteworthy that comprehensive studies exclusively dedicated to DL for multi-label learning remain scarce. Hence, this survey aims to meticulously review recent advancements in DL for multi-label learning while also offering a concise overview of open research issues in MLC. The survey consolidates existing research efforts in DL for MLC, such as deep neural networks, transformers, autoencoders, and convolutional and recurrent architectures. Finally, the study provides a comparative analysis of existing approaches to furnish insightful observations and provoke future research trajectories in this domain.