Thank you for maintaining the comprehensive list of papers on long-tailed learning.
We would like to introduce two recent papers to be considered for inclusion:
Harnessing Hierarchical Label Distribution Variations in Test Agnostic Long-tail Recognition (ICML 2024)[pdf][code]
This paper presents DirMixE, an innovative Mixture-of-Expert framework designed for test-agnostic long-tail recognition. DirMixE utilizes Dirichlet meta-distributions to capture both local and global variations in label distributions, enhancing the framework's generalization and performance stability.
A Unified Generalization Analysis of Re-Weighting and Logit-Adjustment for Imbalanced Learning (NeurIPS 2023)[pdf][code]
This paper introduces a novel data-dependent contraction technique to provide a detailed generalization bound for imbalanced learning. It offers a unified explanation for re-weighting and logit-adjustment methods and proposes a principled learning algorithm validated by empirical results on benchmark datasets.
We would greatly appreciate it if you could include these papers in your list.
Hi,
Thank you for maintaining the comprehensive list of papers on long-tailed learning.
We would like to introduce two recent papers to be considered for inclusion:
Harnessing Hierarchical Label Distribution Variations in Test Agnostic Long-tail Recognition (ICML 2024) [pdf] [code]
This paper presents DirMixE, an innovative Mixture-of-Expert framework designed for test-agnostic long-tail recognition. DirMixE utilizes Dirichlet meta-distributions to capture both local and global variations in label distributions, enhancing the framework's generalization and performance stability.
A Unified Generalization Analysis of Re-Weighting and Logit-Adjustment for Imbalanced Learning (NeurIPS 2023) [pdf] [code]
This paper introduces a novel data-dependent contraction technique to provide a detailed generalization bound for imbalanced learning. It offers a unified explanation for re-weighting and logit-adjustment methods and proposes a principled learning algorithm validated by empirical results on benchmark datasets.
We would greatly appreciate it if you could include these papers in your list.
Thank you very much!