Kwang-Seong Shin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Kwang-Seong Shin has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Seong-Yoon Shin
Department of Computer Information Engineering, Kunsan National University, Gunsan 54150, Republic of Korea
12 shared publications - Sungkwan Youm
Sunchon National University
7 shared publications - Dong-Hyun Won
Office of Liberal Arts Education, Wonkwang University, Iksan 54538, Republic of Korea
5 shared publications - Chuanxiang Song
Shaanxi Institute of Technology, Xi’an 710300, China
4 shared publications - Jong-Chan Kim
Sunchon National University
4 shared publications - Hyuk-Gyu Park
Department of Computer and Software Engineering, Wonkwang University, Iksan 54538, Republic of Korea
3 shared publications
Latest Publications
- Open Access
- 16 November 2026
- Published by MDPI AG in Electronics
- Vol. 15 (22) , 5333
Learning from Noisy Labels in Medical Image Classification: Transition-Matrix Estimation, Loss Correction, and Sample Selection on MedMNIST Benchmarks
Loss corrections based on the noise transition matrix T promise to undo the annotation errors that deep networks memorize in medical images. On six public MedMNIST benchmarks with synthetic label noise, we separately evaluate the estimation of T, the selection of clean examples, and the clean test accuracy. Two propositions show that backward correction inflates the loss variance and that forward correction with an estimate changes decisions only where the clean margin is below twice an explicit error measure of the estimate. In 2175 runs, backward correction was below cross-entropy in 29 of 30 noisy settings, and the cost of forward correction grew with this error measure. The proposed validation-gated forward correction (VGFC), which chooses between forward correction and small-loss selection without clean labels or a known noise rate, was more accurate on average than either candidate but not clearly better than a disagreement-gated variant (+0.13 percentage points, 95% CI [−1.41, +1.24]). Post hoc, it was close to self-adaptive training on the four datasets with accurate estimates of T (above it with ResNet-18 on three of them) and below it on the other two. The benefit of transition-based methods is thus bounded by the accuracy of the estimated matrix. The conclusions are limited to synthetic noise on benchmark data.
- 16 November 2026
- Published by MDPI AG in Electronics
- Vol. 15 (22) , 5333
Open AccessLearning from Noisy Labels in Medical Image Classification: Transition-Matrix Estimation, Loss Correction, and Sample Selection on MedMNIST Benchmarks
Loss corrections based on the noise transition matrix T promise to undo the annotation errors that deep networks memorize in medical images. On six public MedMNIST benchmarks with synthetic label noise, we separately evaluate the estimation of T, the selection of clean examples, and the clean test accuracy. Two propositions show that backward correction inflates the loss variance and that forward correction with an estimate changes decisions only where the clean margin is below twice an explicit error measure of the estimate. In 2175 runs, backward correction was below cross-entropy in 29 of 30 noisy settings, and the cost of forward correction grew with this error measure. The proposed validation-gated forward correction (VGFC), which chooses between forward correction and small-loss selection without clean labels or a known noise rate, was more accurate on average than either candidate but not clearly better than a disagreement-gated variant (+0.13 percentage points, 95% CI [−1.41, +1.24]). Post hoc, it was close to self-adaptive training on the four datasets with accurate estimates of T (above it with ResNet-18 on three of them) and below it on the other two. The benefit of transition-based methods is thus bounded by the accuracy of the estimated matrix. The conclusions are limited to synthetic noise on benchmark data.
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (22) , 55191-55205
FedTQK: Communication-Efficient Quantum Kernel Federated Learning for Anomaly Detection in Resource-Constrained IoT Networks
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (22) , 55191-55205
FedTQK: Communication-Efficient Quantum Kernel Federated Learning for Anomaly Detection in Resource-Constrained IoT Networks
- Open Access
- 27 October 2026
- Published by MDPI AG in Electronics
- Vol. 15 (21) , 4981
Attribute Noise in Tabular Classification: Noise-Matched Training and the Train–Test Noise Mismatch on Public Benchmarks
Attribute noise, the measurement error in feature values, is a data-related challenge distinct from label noise, and the noise present at training time rarely equals the noise present at deployment. This article studies this train–test noise mismatch in tabular classification. Its main contribution is noise-matched training (NMT), which estimates a feature-wise noise level from the variance excess of unlabeled deployment inputs over the training inputs, with a threshold calibrated on the training split, and injects noise of that level in one additional fit. Its identifiability assumption, that the clean deployment variance equals the training variance, is stated and tested. In a one-dimensional Gaussian model, the excess risk of a rule tuned at the wrong noise level vanishes only at the matched level. Five learners on ten public OpenML/UCI datasets were evaluated under eight test-time noise conditions and twelve training conditions, with the dataset as the unit of inference. The matched level was, on average, the best fixed training level, except for nearest-neighbor classification. NMT had the highest average accuracy and helped most when only some features were noisy, but none of its dataset-level comparisons remained significant after multiplicity adjustment. It gave no benefit under impulsive noise, for nearest neighbors, or under covariate shift without noise.
- 27 October 2026
- Published by MDPI AG in Electronics
- Vol. 15 (21) , 4981
Open AccessAttribute Noise in Tabular Classification: Noise-Matched Training and the Train–Test Noise Mismatch on Public Benchmarks
Attribute noise, the measurement error in feature values, is a data-related challenge distinct from label noise, and the noise present at training time rarely equals the noise present at deployment. This article studies this train–test noise mismatch in tabular classification. Its main contribution is noise-matched training (NMT), which estimates a feature-wise noise level from the variance excess of unlabeled deployment inputs over the training inputs, with a threshold calibrated on the training split, and injects noise of that level in one additional fit. Its identifiability assumption, that the clean deployment variance equals the training variance, is stated and tested. In a one-dimensional Gaussian model, the excess risk of a rule tuned at the wrong noise level vanishes only at the matched level. Five learners on ten public OpenML/UCI datasets were evaluated under eight test-time noise conditions and twelve training conditions, with the dataset as the unit of inference. The matched level was, on average, the best fixed training level, except for nearest-neighbor classification. NMT had the highest average accuracy and helped most when only some features were noisy, but none of its dataset-level comparisons remained significant after multiplicity adjustment. It gave no benefit under impulsive noise, for nearest neighbors, or under covariate shift without noise.
- Open Access
- 6 August 2026
- Published by MDPI AG in Electronics
- Vol. 15 (15) , 3488
Frequency–Time Domain Alignment for Cross-Subject Time-Series Classification Under Distribution Shift
Background: Distribution shift is one of the most pervasive data-related challenges for time-series classification. In sensor-based Human Activity Recognition (HAR), a model trained on one population of subjects often degrades substantially when deployed on another, even with identical hardware and activity labels. This paper addresses that setting specifically: cross-subject distribution shift. Methods: We propose FTDA (Frequency–Time Domain Alignment), an unsupervised domain-adaptation method that processes raw signals and their FFT-magnitude spectra through a dual-branch encoder, combines a Gradient-Reversal adversarial loss on the joint feature with a Multi-Kernel MMD on the frequency branch, and applies a symmetric-KL time–frequency consistency loss on unlabeled target data. Results: On UCI HAR, over four cross-subject transfer tasks and 10 random seeds ( paired observations), FTDA reaches 95.01% target accuracy against 92.46% for Source-only, 92.29% for DANN and 93.32% for a RAINCOAT-style baseline. Because the tasks share subjects and two of them are mirror configurations, no pooled significance test is offered; inference is confined to within-task paired tests (10 seeds each, Holm-corrected), under which FTDA is ahead of every single-view baseline on all four tasks, while the difference from the two dual-view competitors ( pp against RAINCOAT-style, descriptive task-level interval ) is not established. Over a common set of seeds, controls widened to FTDA’s parameter and multiply-accumulate budgets shifted accuracy by at most pp against DANN (intervals including zero), where FTDA gains pp on the same runs, so added capacity alone does not appear to explain the improvement, and a target-only reference model, which is optimistic because it is evaluated on the windows it was trained on, sits 0.70 pp above FTDA. An ablation shows that the frequency branch is useful chiefly because it enables a coupled mechanism of cross-view consistency and auxiliary source supervision ( pp for that final step) rather than concatenated features ( pp); for the frequency MMD term, no measurable benefit was detected ( pp, 95% CI ), and it can be removed. On a second dataset, HHAR, the cross-subject advantage does not reproduce under cross-device shift: we detect no difference from time-only adversarial baselines (paired difference pp against DANN, 95% CI ; exploratory, since all device pairs share the same subjects), so the benefit of the spectral view does not appear to carry across shift types. Conclusions: Under cross-subject shift, the operative ingredient is not the spectral view as an extra set of features but the coupled consistency-and-auxiliary mechanism that the second view makes possible; the alignment objectives themselves contribute little, and no measurable benefit was detected for the frequency MMD term. FTDA realizes this view at a modest 0.31 M/7.55 MMAC footprint with no OT solver in the training loop and no hyper-parameter that requires target labels. Transfer across devices and operating environments is not evaluated here and remains open for future work.- 6 August 2026
- Published by MDPI AG in Electronics
- Vol. 15 (15) , 3488
Open AccessFrequency–Time Domain Alignment for Cross-Subject Time-Series Classification Under Distribution Shift
Background: Distribution shift is one of the most pervasive data-related challenges for time-series classification. In sensor-based Human Activity Recognition (HAR), a model trained on one population of subjects often degrades substantially when deployed on another, even with identical hardware and activity labels. This paper addresses that setting specifically: cross-subject distribution shift. Methods: We propose FTDA (Frequency–Time Domain Alignment), an unsupervised domain-adaptation method that processes raw signals and their FFT-magnitude spectra through a dual-branch encoder, combines a Gradient-Reversal adversarial loss on the joint feature with a Multi-Kernel MMD on the frequency branch, and applies a symmetric-KL time–frequency consistency loss on unlabeled target data. Results: On UCI HAR, over four cross-subject transfer tasks and 10 random seeds ( paired observations), FTDA reaches 95.01% target accuracy against 92.46% for Source-only, 92.29% for DANN and 93.32% for a RAINCOAT-style baseline. Because the tasks share subjects and two of them are mirror configurations, no pooled significance test is offered; inference is confined to within-task paired tests (10 seeds each, Holm-corrected), under which FTDA is ahead of every single-view baseline on all four tasks, while the difference from the two dual-view competitors ( pp against RAINCOAT-style, descriptive task-level interval ) is not established. Over a common set of seeds, controls widened to FTDA’s parameter and multiply-accumulate budgets shifted accuracy by at most pp against DANN (intervals including zero), where FTDA gains pp on the same runs, so added capacity alone does not appear to explain the improvement, and a target-only reference model, which is optimistic because it is evaluated on the windows it was trained on, sits 0.70 pp above FTDA. An ablation shows that the frequency branch is useful chiefly because it enables a coupled mechanism of cross-view consistency and auxiliary source supervision ( pp for that final step) rather than concatenated features ( pp); for the frequency MMD term, no measurable benefit was detected ( pp, 95% CI ), and it can be removed. On a second dataset, HHAR, the cross-subject advantage does not reproduce under cross-device shift: we detect no difference from time-only adversarial baselines (paired difference pp against DANN, 95% CI ; exploratory, since all device pairs share the same subjects), so the benefit of the spectral view does not appear to carry across shift types. Conclusions: Under cross-subject shift, the operative ingredient is not the spectral view as an extra set of features but the coupled consistency-and-auxiliary mechanism that the second view makes possible; the alignment objectives themselves contribute little, and no measurable benefit was detected for the frequency MMD term. FTDA realizes this view at a modest 0.31 M/7.55 MMAC footprint with no OT solver in the training loop and no hyper-parameter that requires target labels. Transfer across devices and operating environments is not evaluated here and remains open for future work. - Open Access
- 4 August 2026
- Published by MDPI AG in Electronics
- Vol. 15 (15) , 3447
Confidence-Calibrated Consistency Matching for Semi-Supervised Image Classification Under Extreme Label Scarcity
Labeling images is expensive, but unlabeled data is abundant. Semi-supervised learning (SSL) addresses this gap, though the dominant pseudo-labeling methods can suffer from confirmation bias—reinforcing their own confident-but-wrong predictions—most severely when labels are scarcest. Under a controlled, reproducible compute-constrained protocol on CIFAR-10, SVHN, and CIFAR-100, we examine which ingredients of consistency-based SSL actually help when as few as four labels per class are available. We propose CCM (Confidence-Calibrated Consistency Matching)—a per-class curriculum threshold, a dual strong-view consistency loss, and a smooth confidence weighting that softly admits borderline pseudo-labels—together with a unified view in which FixMatch, FlexMatch, SoftMatch, and CCM instantiate a single generalized weighting function. On CIFAR-10 with 40 labels, CCM reaches 35.14%, a significant improvement over the FlexMatch design it directly extends (+3.30 percentage points (pp), paired t-test p = 0.001). SoftMatch, re-trained under the identical budget, performs better still at the two smallest budgets (37.04% at 40 labels, p = 0.049), while CCM leads numerically at 4000 labels: the two smooth-weighting designs top the extreme-scarcity board—convergent evidence that the smoothness of the weighting function, more than the placement of its threshold, is the decisive design axis. We also report a negative result: cross-view agreement helps neither as an admission gate nor as reliability reweighting, reducing accuracy by up to 4.37 pp; agreement is a positive correctness signal, but its absolute level (approximately 44% correct among agreeing pseudo-labels) is too low to filter on safely. Curriculum thresholding, by contrast, hurts on the easier SVHN dataset and fails outright on CIFAR-100 when its per-class statistics become too thin. CCM adds no inference-time cost. We do not claim universality; we characterize when each ingredient helps within a single, identical-budget protocol.
- 4 August 2026
- Published by MDPI AG in Electronics
- Vol. 15 (15) , 3447
Open AccessConfidence-Calibrated Consistency Matching for Semi-Supervised Image Classification Under Extreme Label Scarcity
Labeling images is expensive, but unlabeled data is abundant. Semi-supervised learning (SSL) addresses this gap, though the dominant pseudo-labeling methods can suffer from confirmation bias—reinforcing their own confident-but-wrong predictions—most severely when labels are scarcest. Under a controlled, reproducible compute-constrained protocol on CIFAR-10, SVHN, and CIFAR-100, we examine which ingredients of consistency-based SSL actually help when as few as four labels per class are available. We propose CCM (Confidence-Calibrated Consistency Matching)—a per-class curriculum threshold, a dual strong-view consistency loss, and a smooth confidence weighting that softly admits borderline pseudo-labels—together with a unified view in which FixMatch, FlexMatch, SoftMatch, and CCM instantiate a single generalized weighting function. On CIFAR-10 with 40 labels, CCM reaches 35.14%, a significant improvement over the FlexMatch design it directly extends (+3.30 percentage points (pp), paired t-test p = 0.001). SoftMatch, re-trained under the identical budget, performs better still at the two smallest budgets (37.04% at 40 labels, p = 0.049), while CCM leads numerically at 4000 labels: the two smooth-weighting designs top the extreme-scarcity board—convergent evidence that the smoothness of the weighting function, more than the placement of its threshold, is the decisive design axis. We also report a negative result: cross-view agreement helps neither as an admission gate nor as reliability reweighting, reducing accuracy by up to 4.37 pp; agreement is a positive correctness signal, but its absolute level (approximately 44% correct among agreeing pseudo-labels) is too low to filter on safely. Curriculum thresholding, by contrast, hurts on the easier SVHN dataset and fails outright on CIFAR-100 when its per-class statistics become too thin. CCM adds no inference-time cost. We do not claim universality; we characterize when each ingredient helps within a single, identical-budget protocol.
- Open Access
- 30 June 2026
- Published by Korea Institute of Information and Communication Engineering in Journal of information and communication convergence engineering
- Vol. 24 (2) , 143-154
Integrated Analysis of Missing-Value Imputation Techniques and Class-Imbalance Handling for CKD Prediction
- 30 June 2026
- Published by Korea Institute of Information and Communication Engineering in Journal of information and communication convergence engineering
- Vol. 24 (2) , 143-154
Open AccessIntegrated Analysis of Missing-Value Imputation Techniques and Class-Imbalance Handling for CKD Prediction
- Top Cited PapersOpen Access
- 20 February 2026
- Published by MDPI AG in Electronics
- Vol. 15 (4) , 883
Synthetic Data Augmentation for Imbalanced Tabular Data: A Comparative Study of Generation Methods
Class imbalance in tabular datasets poses a challenge for machine learning classification tasks, often leading to biased models that underperform in predicting minority class instances. This study presents a comparative analysis of synthetic data generation methods for addressing class imbalance in tabular data. We evaluate four augmentation approaches—Synthetic Minority Over-sampling Technique (SMOTE), Gaussian Copula, Tabular Variational Autoencoder (TVAE), and Conditional Tabular Generative Adversarial Network (CTGAN)—using the University of California Irvine (UCI) Bank Marketing dataset, which exhibits a class imbalance ratio of approximately 7.88:1. Our experimental framework assesses each method across three dimensions: statistical fidelity to the original data distribution evaluated through four complementary metrics (marginal numerical similarity, categorical distribution similarity, correlation structure preservation, and Kolmogorov–Smirnov test), machine learning utility measured through classification performance, and minority class detection capability. Results indicate that all augmentation methods achieved statistically significant improvements over the baseline (). SMOTE achieved the highest recall (54.2%, a 117.6% relative improvement over the baseline) and F1-Score (0.437, +22.4% over the baseline) for minority class detection, while Gaussian Copula provided the highest composite fidelity score (0.930) with competitive predictive performance. A weak negative correlation () between composite fidelity and classification performance was observed, suggesting that higher statistical fidelity does not necessarily translate to better downstream task performance. Deep learning-based methods (TVAE, CTGAN) showed statistically significant improvements over the baseline (recall: +58% to +63%) but underperformed compared to simpler methods under default configurations, suggesting the need for larger training samples or more extensive hyperparameter tuning. These findings offer reference points for practitioners working with moderately imbalanced tabular data with limited minority class samples, supporting the selection of generation strategies based on specific requirements regarding data fidelity and classification objectives.- 20 February 2026
- Published by MDPI AG in Electronics
- Vol. 15 (4) , 883
Top Cited PapersOpen AccessSynthetic Data Augmentation for Imbalanced Tabular Data: A Comparative Study of Generation Methods
Class imbalance in tabular datasets poses a challenge for machine learning classification tasks, often leading to biased models that underperform in predicting minority class instances. This study presents a comparative analysis of synthetic data generation methods for addressing class imbalance in tabular data. We evaluate four augmentation approaches—Synthetic Minority Over-sampling Technique (SMOTE), Gaussian Copula, Tabular Variational Autoencoder (TVAE), and Conditional Tabular Generative Adversarial Network (CTGAN)—using the University of California Irvine (UCI) Bank Marketing dataset, which exhibits a class imbalance ratio of approximately 7.88:1. Our experimental framework assesses each method across three dimensions: statistical fidelity to the original data distribution evaluated through four complementary metrics (marginal numerical similarity, categorical distribution similarity, correlation structure preservation, and Kolmogorov–Smirnov test), machine learning utility measured through classification performance, and minority class detection capability. Results indicate that all augmentation methods achieved statistically significant improvements over the baseline (). SMOTE achieved the highest recall (54.2%, a 117.6% relative improvement over the baseline) and F1-Score (0.437, +22.4% over the baseline) for minority class detection, while Gaussian Copula provided the highest composite fidelity score (0.930) with competitive predictive performance. A weak negative correlation () between composite fidelity and classification performance was observed, suggesting that higher statistical fidelity does not necessarily translate to better downstream task performance. Deep learning-based methods (TVAE, CTGAN) showed statistically significant improvements over the baseline (recall: +58% to +63%) but underperformed compared to simpler methods under default configurations, suggesting the need for larger training samples or more extensive hyperparameter tuning. These findings offer reference points for practitioners working with moderately imbalanced tabular data with limited minority class samples, supporting the selection of generation strategies based on specific requirements regarding data fidelity and classification objectives. - Open Access
- 3 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 22545-22568
Force Feedback-Based Adaptive Sensor Fusion Algorithm for Precision Enhancement in VR-Based Medical Training Systems
Sub-millimeter tracking accuracy is an important requirement for high-fidelity simulation in Virtual Reality (VR)-based medical training systems. Conventional optical tracking systems suffer from significant errors during occlusion, while standalone Inertial Measurement Unit (IMU) usage leads to severe drift accumulation. This study proposes an Adaptive Extended Kalman Filter (AEKF)-based sensor fusion algorithm that integrates optical tracking, IMU, and force/torque sensors. The key technical contributions are: (1) a force feedback-based position drift correction mechanism, (2) innovation-based adaptive process noise adjustment, and (3) real-time tissue stiffness estimation through a 21-dimensional state vector. To evaluate the proposed algorithm’s performance, we generated simulation datasets for three standard medical scenarios (laparoscopic suturing, catheter insertion, and needle biopsy) and quantitatively compared five sensor fusion algorithms: (A) optical-only, (B) IMU-only, (C) standard EKF (optical+IMU), (D) Unscented KF (optical+IMU), and (E) proposed method (optical+IMU+force). Simulation-based experiments show that the proposed algorithm achieves an average position tracking error of 0.75mm (±0.19mm) in controlled simulation conditions, representing 95.5% improvement over optical-only (P < 0.05) and 85.4% improvement over standard EKF (p < 0.01) within the simulation environment. Furthermore, it maintains stable performance below 1.2mm even during simulated optical occlusion and achieves computational latency of 2.1ms on the tested hardware platform. This research focuses on algorithmic development and simulation-based validation, and does not include hardware prototype implementation, user studies with medical professionals, or clinical validation. The results demonstrate algorithmic feasibility under idealized simulation conditions rather than validated performance in real clinical environments.
- 3 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 22545-22568
Open AccessForce Feedback-Based Adaptive Sensor Fusion Algorithm for Precision Enhancement in VR-Based Medical Training Systems
Sub-millimeter tracking accuracy is an important requirement for high-fidelity simulation in Virtual Reality (VR)-based medical training systems. Conventional optical tracking systems suffer from significant errors during occlusion, while standalone Inertial Measurement Unit (IMU) usage leads to severe drift accumulation. This study proposes an Adaptive Extended Kalman Filter (AEKF)-based sensor fusion algorithm that integrates optical tracking, IMU, and force/torque sensors. The key technical contributions are: (1) a force feedback-based position drift correction mechanism, (2) innovation-based adaptive process noise adjustment, and (3) real-time tissue stiffness estimation through a 21-dimensional state vector. To evaluate the proposed algorithm’s performance, we generated simulation datasets for three standard medical scenarios (laparoscopic suturing, catheter insertion, and needle biopsy) and quantitatively compared five sensor fusion algorithms: (A) optical-only, (B) IMU-only, (C) standard EKF (optical+IMU), (D) Unscented KF (optical+IMU), and (E) proposed method (optical+IMU+force). Simulation-based experiments show that the proposed algorithm achieves an average position tracking error of 0.75mm (±0.19mm) in controlled simulation conditions, representing 95.5% improvement over optical-only (P < 0.05) and 85.4% improvement over standard EKF (p < 0.01) within the simulation environment. Furthermore, it maintains stable performance below 1.2mm even during simulated optical occlusion and achieves computational latency of 2.1ms on the tested hardware platform. This research focuses on algorithmic development and simulation-based validation, and does not include hardware prototype implementation, user studies with medical professionals, or clinical validation. The results demonstrate algorithmic feasibility under idealized simulation conditions rather than validated performance in real clinical environments.
- Open Access
- 10 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 210010-210024
Theoretically Grounded Adaptive Sampling for Imbalanced Multi-Label Classification in Medical Text
Multi-label medical text classification faces severe class imbalance, where rare pathological conditions appear hundreds of times less frequently than common findings. This imbalance causes standard neural networks to achieve high overall accuracy while failing to detect rare but clinically important conditions. Existing sampling strategies rely on heuristic rules rather than theoretical foundations and treat all minority classes uniformly despite their varying degrees of rarity. We propose two adaptive sampling strategies: a rule-based method using percentile-based categorization and a formula-based method grounded in Vapnik-Chervonenkis (VC) dimension theory. The formula-based approach derives upsampling weights through wc = ln(rc + 1) · αc, combining logarithmic scaling of imbalance ratios with smooth interpolation. This formulation aligns upsampling decisions with sample complexity requirements from statistical learning theory while preventing overfitting. Experiments on the PadChest chest X-ray report dataset demonstrate that the formula-based method achieves 31.9% improvement in Macro F1 with 34.5% fewer training samples than rule-based sampling. The method reduces the F1 gap from 0.51 to 0.42, demonstrating significantly improved balance between rare and common class performance. Results validate that theoretically-motivated sampling strategies yield measurable benefits over heuristic approaches.
- 10 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 210010-210024
Open AccessTheoretically Grounded Adaptive Sampling for Imbalanced Multi-Label Classification in Medical Text
Multi-label medical text classification faces severe class imbalance, where rare pathological conditions appear hundreds of times less frequently than common findings. This imbalance causes standard neural networks to achieve high overall accuracy while failing to detect rare but clinically important conditions. Existing sampling strategies rely on heuristic rules rather than theoretical foundations and treat all minority classes uniformly despite their varying degrees of rarity. We propose two adaptive sampling strategies: a rule-based method using percentile-based categorization and a formula-based method grounded in Vapnik-Chervonenkis (VC) dimension theory. The formula-based approach derives upsampling weights through wc = ln(rc + 1) · αc, combining logarithmic scaling of imbalance ratios with smooth interpolation. This formulation aligns upsampling decisions with sample complexity requirements from statistical learning theory while preventing overfitting. Experiments on the PadChest chest X-ray report dataset demonstrate that the formula-based method achieves 31.9% improvement in Macro F1 with 34.5% fewer training samples than rule-based sampling. The method reduces the F1 gap from 0.51 to 0.42, demonstrating significantly improved balance between rare and common class performance. Results validate that theoretically-motivated sampling strategies yield measurable benefits over heuristic approaches.