This repository is a curated collection of academic papers and resources focused on "long-tail learning," a machine learning paradigm addressing datasets with highly imbalanced label distributions. It serves researchers and practitioners in computer vision and natural language processing who encounter or aim to mitigate the challenges posed by such data skew, particularly in image classification and extreme multi-label learning.

How It Works

The repository categorizes papers based on specific long-tail learning sub-problems, including semi-supervised learning, noisy labels, out-of-distribution detection, and federated learning. It also extensively covers extreme multi-label learning (XML) with sub-categories like binary relevance, tree-based methods, and embedding-based approaches. The organization facilitates a structured understanding of the research landscape and common methodologies such as Two-Stage Training (TST), Instance Sampling (IS), Class-Balanced Sampling (CBS), Class-Level Weighting (CLW), Normalized Classifier (NC), Ensemble methods (ENS), and Data Augmentation (DA).

Quick Start & Requirements

This is a curated list of papers and does not involve direct code execution or installation. The primary requirement is access to academic literature and potentially the linked code repositories for individual papers.

Highlighted Details

• Comprehensive coverage of long-tail learning applications in computer vision (image classification) and extreme multi-label learning (text categorization).
• Detailed categorization of papers by sub-problem and methodology, including links to associated code where available.
• Extensive lists of papers with publication venues, years, and brief remarks, updated as of July 2024.
• Includes links to relevant workshops, seminars, and survey papers for deeper exploration.

Maintenance & Community

The repository is maintained by weitongseu and was last updated on 2024-07-13, indicating active curation. Specific community channels or contributor details beyond the maintainer are not provided.

Licensing & Compatibility

The repository itself is a collection of links and summaries; it does not have a specific license. The licensing of individual papers and their associated code would need to be checked on a per-paper basis.

Limitations & Caveats

This repository is a literature aggregator and does not provide a unified codebase or framework for long-tail learning. Users must refer to individual papers for implementation details and code.