Generalizable Brain Tumor Segmentation with Self-Training and Tumor-Aware Deformations
该研究针对脑肿瘤分割的泛化问题，采用nnU-Net框架结合半监督学习和肿瘤感知变形增强方法，提高了不同肿瘤区域的分割精度。
0 citationsRead paper
Selected work
Representative Papers
An Ensemble-Based Self-Taught Learning Approach for Parking Space Classification Under Limited Data
本文针对停车空间分类数据不足的问题，提出了一种基于自学习框架和卷积自动编码器的方法，通过无监督表示学习及异构自动编码器集成提高分类准确性和鲁棒性。
0 citationsRead paper
Hoeffding adaptive splitting trees for data stream classification with concept drift and ensemble learning
This study addresses the insufficient diversity of adaptive splitting trees when employed as base learners in ensembles for data stream classification. We propose the Hoeffding Adaptive Splitting Tree, which integrates periodic splitting strategies with adaptive mechanisms. By leveraging change detection to precisely identify performance degradation and determine optimal split points, this approach effectively enhances ensemble diversity. Experimental results demonstrate that the proposed model achieves an optimal trade-off between classification accuracy and computational efficiency under concept drift scenarios. Furthermore, it attains state-of-the-art performance across benchmark evaluations, computational cost analyses, and drift adaptability assessments, thereby providing an efficient solution for learning from streaming data.
0 citationsRead paper
Computer Vision for MOBA Analytics: A Dataset and Baseline for Visibility Analysis in Dota 2
This study addresses the limitation of existing MOBA game analyses, which predominantly rely on structured data and fail to capture the actual in-game visibility available to teams during matches. To bridge this gap, the authors introduce Dota2-Vis, a novel video dataset, and propose the first visibility analysis framework based on dual-perspective gameplay footage and manually annotated minimap images. Leveraging the YOLOv11 model family, they process 288 full-HD match videos and 2,477 minimap images to infer the presence states of opposing players. Experimental results demonstrate that YOLOv11l achieves superior performance in handling dense and cluttered minimap scenes, generating highly reliable visibility curves. These curves effectively reveal behavioral patterns at the levels of individual players, heroes, roles, and entire teams, thereby complementing and extending traditional structured-data approaches.
0 citationsRead paper
Univariate Channel Fusion for Multivariate Time Series Classification
This work addresses the high computational cost of multivariate time series classification (MTSC) models, which hinders their real-time deployment on resource-constrained devices. To overcome this limitation, the authors propose a lightweight channel fusion mechanism that compresses multivariate sequences into univariate representations using strategies such as mean, median, or dynamic time warping (DTW) barycenter aggregation, thereby enabling compatibility with efficient off-the-shelf univariate classifiers. The approach proves particularly effective when channels exhibit strong inter-variable correlations. Evaluated across five diverse datasets spanning chemical monitoring, brain–computer interfaces, and human activity recognition, the method not only significantly outperforms existing MTSC baselines and state-of-the-art approaches but also achieves substantial reductions in computational complexity, effectively balancing high accuracy with high efficiency.
0 citationsRead paper
Recent publications
Latest Papers
Generalizable Brain Tumor Segmentation with Self-Training and Tumor-Aware Deformations
该研究针对脑肿瘤分割的泛化问题，采用nnU-Net框架结合半监督学习和肿瘤感知变形增强方法，提高了不同肿瘤区域的分割精度。
0 citationsRead paper
An Ensemble-Based Self-Taught Learning Approach for Parking Space Classification Under Limited Data
本文针对停车空间分类数据不足的问题，提出了一种基于自学习框架和卷积自动编码器的方法，通过无监督表示学习及异构自动编码器集成提高分类准确性和鲁棒性。
0 citationsRead paper
Hoeffding adaptive splitting trees for data stream classification with concept drift and ensemble learning
This study addresses the insufficient diversity of adaptive splitting trees when employed as base learners in ensembles for data stream classification. We propose the Hoeffding Adaptive Splitting Tree, which integrates periodic splitting strategies with adaptive mechanisms. By leveraging change detection to precisely identify performance degradation and determine optimal split points, this approach effectively enhances ensemble diversity. Experimental results demonstrate that the proposed model achieves an optimal trade-off between classification accuracy and computational efficiency under concept drift scenarios. Furthermore, it attains state-of-the-art performance across benchmark evaluations, computational cost analyses, and drift adaptability assessments, thereby providing an efficient solution for learning from streaming data.
0 citationsRead paper
Computer Vision for MOBA Analytics: A Dataset and Baseline for Visibility Analysis in Dota 2
This study addresses the limitation of existing MOBA game analyses, which predominantly rely on structured data and fail to capture the actual in-game visibility available to teams during matches. To bridge this gap, the authors introduce Dota2-Vis, a novel video dataset, and propose the first visibility analysis framework based on dual-perspective gameplay footage and manually annotated minimap images. Leveraging the YOLOv11 model family, they process 288 full-HD match videos and 2,477 minimap images to infer the presence states of opposing players. Experimental results demonstrate that YOLOv11l achieves superior performance in handling dense and cluttered minimap scenes, generating highly reliable visibility curves. These curves effectively reveal behavioral patterns at the levels of individual players, heroes, roles, and entire teams, thereby complementing and extending traditional structured-data approaches.
0 citationsRead paper
Univariate Channel Fusion for Multivariate Time Series Classification
This work addresses the high computational cost of multivariate time series classification (MTSC) models, which hinders their real-time deployment on resource-constrained devices. To overcome this limitation, the authors propose a lightweight channel fusion mechanism that compresses multivariate sequences into univariate representations using strategies such as mean, median, or dynamic time warping (DTW) barycenter aggregation, thereby enabling compatibility with efficient off-the-shelf univariate classifiers. The approach proves particularly effective when channels exhibit strong inter-variable correlations. Evaluated across five diverse datasets spanning chemical monitoring, brain–computer interfaces, and human activity recognition, the method not only significantly outperforms existing MTSC baselines and state-of-the-art approaches but also achieves substantial reductions in computational complexity, effectively balancing high accuracy with high efficiency.
0 citationsRead paper