![[2509.00955] ART: Adaptive Resampling-based Training for Imbalanced Classification](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2509.00955] ART: Adaptive Resampling-based Training for Imbalanced Classification
Summary
The paper presents ART, a novel Adaptive Resampling-based Training method for imbalanced classification that dynamically adjusts training data distribution based on class performance, leading to improved model accuracy.
Why It Matters
Imbalanced classification is a common challenge in machine learning, often leading to suboptimal model performance. ART addresses this issue by adapting the training process based on real-time performance metrics, making it a significant advancement over traditional static methods. This approach can enhance the reliability of models in critical applications across various domains.
Key Takeaways
- ART adapts training data distribution based on class performance metrics.
- The method outperforms traditional resampling techniques like SMOTE and NearMiss.
- Improvements in macro F1 score are statistically significant across multiple datasets.
- ART provides a consistent performance boost in both binary and multi-class classification tasks.
- The approach is robust against varying degrees of class imbalance.
Computer Science > Machine Learning arXiv:2509.00955 (cs) [Submitted on 31 Aug 2025 (v1), last revised 15 Feb 2026 (this version, v2)] Title:ART: Adaptive Resampling-based Training for Imbalanced Classification Authors:Arjun Basandrai, Shourya Jain, K. Ilanthenral View a PDF of the paper titled ART: Adaptive Resampling-based Training for Imbalanced Classification, by Arjun Basandrai and 1 other authors View PDF HTML (experimental) Abstract:Traditional resampling methods for handling class imbalance typically uses fixed distributions, undersampling the majority or oversampling the minority. These static strategies ignore changes in class-wise learning difficulty, which can limit the overall performance of the model. This paper proposes an Adaptive Resampling-based Training (ART) method that periodically updates the distribution of the training data based on the class-wise performance of the model. Specifically, ART uses class-wise macro F1 scores, computed at fixed intervals, to determine the degree of resampling to be performed. Unlike instance-level difficulty modeling, which is noisy and outlier-sensitive, ART adapts at the class level. This allows the model to incrementally shift its attention towards underperforming classes in a way that better aligns with the optimization objective. Results on diverse benchmarks, including Pima Indians Diabetes and Yeast dataset demonstrate that ART consistently outperforms both resampling-based and algorithm-level methods, including ...
