Real-world datasets often exhibit severe class imbalance, degrading the performance of deep learning models, particularly on rare classes. While imbalance is well studied in binary and multiclass settings, it is substantially more challenging in multilabel problems due to complex label co-occurrence. In clinical applications such as ECG classification, this challenge is coupled with a second critical challenge: the need to prioritize high-risk conditions, regardless of their prevalence, to mitigate the high cost of false negatives. We introduce the Multilabel Class-Prioritized Probabilistic Sampler (ML-CLAPPS), a learnable sampler that optimizes a distribution over training instances to control the model's exposure to different classes. This approach addresses multilabel imbalance while enabling prioritization of significant conditions via a user-defined importance vector (IV). Empirical evaluation on a real-world multilabel ECG image dataset demonstrates that ML-CLAPPS achieves state-of-the-art performance. We show that it aligns class exposure with the specified IV, and that increasing a class's importance leads to corresponding improvements in its classification performance, enabling explicit prioritization of clinically critical conditions.
Figures & tables
Figure 1: Class distribution of the Mobile dataset, illustrating dataset imbalance.
Figure 2: Optimal p⋆ (left) and d (right) from ML-CLAPPS under varying γ with uniform IV. Larger γ improves dataset utilization via more uniform sampling, at the cost of shifting exposure toward the original imbalanced label distribution (Fig. 1 ).
Figure 3: Effect of γ on dataset utilization loss (left axis) and alignment loss (right axis), derived from solving ML-CLAPPS for each γ under a uniform IV. Lower values denote improved utilization and closer exposure-importance alignment.
Figure 4: Expected class exposure d obtained by solving ML-CLAPPS for different IVs using γ=103 . Top: descending importance. Bottom: alternating importance. Expected exposure closely follows the prescribed class priorities, mirroring the user-defined IVs.
Method
Classification Performance
Overhead (sec) ↓
F1 ↑
MCC ↑
Accuracy ↑
Init
Per-Epoch
UniformSampler (Baseline)
0.281 ± .02
0.301 ± .02
0.954 ± .001
0
0
LP ROS [ 6 , 13 ]
0.291 ± .01
0.315 ± .01
0.949 ± .001
0.05 ± .01
0
ClusterClip Sampling [ 24 ]
0.299 ± .02
0.344 ± .03
0.953 ± .002
0.01 ± .00
0.01 ± .00
LeastSampledClass [ 19 ]
0.325 ± .02
0.365 ± .02
0.954 ± .001
0
0.49 ± .05
ML-CLAPPS (IV= LinR2F )
0.347 ± .01
0.363 ± .01
0.954 ± .001
1.17 ± .20
0
Table 1: Comparative performance of sampling strategies on the Mobile ECG dataset. ML-CLAPPS outperforms prior methods. Table reports mean ± SD across four runs.
Figure 5: The x-axis shows class prevalence from rare (left) to frequent (right), where Rk indicates the kth most rare class. The y-axis reports mean F1 across four runs. For each sampling method, a fitted linear regression line is shown to highlight performance trends across class frequencies. LinR2F IV demonstrates targeted condition prioritization.
The complex imbalanced label distribution poses a crucial challenge to multi-label classification, as most classifiers are biased towards the majority class and high-frequent labels. Oversampling is an efficient and flexible solution that augments instances to provide a more balanced training dataset for multi-label classifiers. Most existing oversampling methods create synthetic instances in a heuristic way that essentially relies on neighborhood information retrieved using Euclidean distance within the entire feature space. However, they fail to consider the varying semantic relevance of features to different labels, leading to label inconsistency among proximate neighbors and further introducing label confusion and overfitting to synthetic instances. To overcome the above issue, we propose a novel sampling approach called Label-Specific Distance-based Multi-Label Oversampling (LSDMLO) that creates more useful and well-labeled synthetic instances to address the imbalance in multi-label datasets. LSDMLO derives the label-specific distance to identify label-consistent neighbors based on the weighted pertinent feature space, which facilitates selecting seed instances that express more label correlations in boundary areas and generating synthetic instances aligned with the label distribution of original data. The comprehensive experiments verify that the proposed LSDMLO outperforms the state-of-the-art multi-label sampling approaches under various base classifiers.
Bin Liu, Jun Wu, Haoyu Peng +4
Key Laboratory of Data Engineering and Visual Computing, Chongqing University of Posts and Telecommunications, China · School of Computer Science and Technology, Chongqing University of Posts and Telecommunications, China · State Key Laboratory of Novel Software Technology, Nanjing University, China +1
Numerous studies on Boosting attempt to mitigate classification bias caused by class imbalance. However, existing studies have yet to explore the collaborative optimization of imbalanced learning and model training. This constraint hinders further performance improvements. To bridge this gap, this study proposes a collaborative optimization Boosting model of multiclass imbalanced learning. By integrating the density factor and the confidence factor, this model implements a noise-resistant weight update mechanism alongside a dynamic sampling strategy. Rather than functioning as independent components, these modules are tightly integrated to orchestrate weight updates, sample region partitioning, and region-guided sampling. Thus, this study proposes the collaborative optimization of imbalanced learning and model training. Extensive experiments on 40 public imbalanced datasets demonstrate that the proposed model significantly outperforms seven state-of-the-art baselines. The code and datasets for this paper are available at: https://github.com/ChuantaoLi/DARG.
Chuantao Li, Zhi Li, Jiahao Xu +2
School of Automation Engineering, University of Electronic Science and Technology of China, Chengdu 611731, Sichuan, China · School of Mathematics and Computer, Guangdong Ocean University, Zhanjiang 524088, Guangdong, China
Multiple Instance Learning (MIL) has emerged as a promising paradigm for Whole Slide Image (WSI) diagnosis, offering effective learning with limited annotations. However, existing MIL frameworks overlook diagnostic priorities and fail to differentiate the severity of misclassifications in multiclass, leaving clinically critical errors unaddressed. We propose a mistake-severity-aware training strategy that organizes diagnostic classes into a hierarchical structure, with each level optimized using a severity-weighted cross-entropy loss that penalizes high-severity misclassifications more strongly. Additionally, hierarchical consistency is enforced through probabilistic alignment, a semantic feature remix applied to the instance bag to robustly train class priority and accommodate clinical cases involving multiple symptoms. An asymmetric Mikel's Wheel-based metric is also introduced to quantify the severity of errors specific to medical fields. Experiments on challenging public and real-world in-house datasets demonstrate that our approach significantly mitigates critical errors in MIL diagnosis compared to existing methods. We present additional experimental results on natural domain data to demonstrate the generalizability of our proposed method beyond medical contexts.
Sungrae Hong, Jiwon Jeong, Jisu Shin +4
Korea Advanced Institute of Science and Technology, Daejeon, South Korea · Seegene Medical Foundation, Seoul, South Korea