Optimizing Class Distributions for Bias-Aware Multi-Class Learning
Not on the map. Why?
Abstract (arXiv)
We propose BiCDO (Bias-Controlled Class Distribution Optimizer), an iterative, data-centric framework that identifies Pareto optimized class distributions for multi-class image classification. BiCDO enables performance prioritization for specific classes, which is useful in safety-critical scenarios (e.g. prioritizing 'Human' over 'Dog'). Unlike uniform distributions, BiCDO determines the optimal number of images per class to enhance reliability and minimize bias and variance in the objective function. BiCDO can be incorporated into existing training pipelines with minimal code changes and supports any labelled multi-class dataset. We have validated BiCDO using EfficientNet, ResNet and ConvNeXt on CIFAR-10 and iNaturalist21 datasets, demonstrating improved, balanced model performance through optimized data distribution.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 34 of its 46 references that are in the map. First posted on arXiv 15 Sep 2025.
Builds on
34 of 46 references.
References · 34
- 2015Deep Residual Learning for Image Recognitioncited by 52,096
- 2019EfficientNet: Rethinking Model Scaling for Convolutional Neural Networkscited by 6,442
- 2022Generative Adversarial Networkscited by 5,627
- 2022A ConvNet for the 2020scited by 3,579
- 2019Learning Imbalanced Datasets with Label-Distribution-Aware Margin Losscited by 989
- 2020Equalization Loss for Long-Tailed Object Recognitioncited by 285
- 2021Improving Calibration for Long-Tailed Recognitioncited by 212
- 2023Data-centric Artificial Intelligence: A Surveycited by 203
- 2021Distribution Alignment: A Unified Framework for Long-tail Visual Recognitioncited by 198
- 2020Disentangling Label Distribution for Long-tailed Visual Recognitioncited by 172
- 2021The Majority Can Help The Minority: Context-rich Minority Oversampling for Long-tailed Classificationcited by 100
- 2021MetaSAug: Meta Semantic Augmentation for Long-Tailed Visual Recognitioncited by 99
22 more
- 2021Self-Supervised Aggregation of Diverse Experts for Test-Agnostic Long-Tailed Recognitioncited by 95
- 2021Distributional Robustness Loss for Long-tail Learningcited by 84
- 2021FASA: Feature Augmentation and Sampling Adaptation for Long-Tailed Instance Segmentationcited by 76
- 2023Long-tailed Visual Recognition via Gaussian Clouded Logit Adjustmentcited by 75
- 2022Auto-Lambda: Disentangling Dynamic Task Relationshipscited by 73
- 2021Distilling Virtual Examples for Long-tailed Recognitioncited by 72
- 2021Learning of Visual Relations: The Devil is in the Tailscited by 65
- 2021RSG: A Simple but Effective Module for Learning Imbalanced Datasetscited by 61
- 2021GistNet: a Geometric Structure Transfer Network for Long-Tailed Recognitioncited by 31
- 2022Data-Centric Artificial Intelligencecited by 31
- 2023Subclass-balancing Contrastive Learning for Long-tailed Recognitioncited by 26
- 2021Imagine by Reasoning: A Reasoning-Based Implicit Semantic Data Augmentation for Long-Tailed Classificationcited by 24
- 2020Class-Wise Difficulty-Balanced Loss for Solving Class-Imbalancecited by 23
- 2022Class-Difficulty Based Methods for Long-Tailed Visual Recognitioncited by 23
- 2022Delving into Semantic Scale Imbalancecited by 22
- 2023No One Left Behind: Improving the Worst Categories in Long-Tailed Learningcited by 14
- 2023Predicting and Enhancing the Fairness of DNNs with the Curvature of Perceptual Manifoldscited by 9
- 2018Clustering and Learning from Imbalanced Datacited by 7
- 2021Exact Pareto Optimal Search for Multi-Task Learning and Multi-Criteria Decision-Makingcited by 6
- 2024Model-Based Data-Centric AI: Bridging the Divide Between Academic Ideals and Industrial Pragmatismcited by 1
- 2024Covariance-corrected Whitening Alleviates Network Degeneration on Imbalanced Classificationcited by 1
- 2024Learn2Mix: Training Neural Networks Using Adaptive Data Integrationcited by 1
Built on by
Not cited yet.