Selected work

Representative Papers

Busemann energy-based attention for emotion analysis in Poincaré discs

Apr 08, 2026

Traditional sentiment analysis treats emotions as discrete categories, failing to capture their semantic hierarchy and inherent ambiguity. This work proposes EmBolic, the first end-to-end trainable fully hyperbolic sentiment analysis architecture that integrates hyperbolic geometry with Busemann energy. Operating within the Poincaré disk model, EmBolic employs a Busemann energy–driven attention mechanism to map input text to query points in hyperbolic space and automatically generates boundary key points to represent emotional directions. The model can infer the curvature of the continuous emotion space, achieving high accuracy and strong generalization even with low-dimensional embeddings. Experimental results demonstrate that EmBolic significantly outperforms conventional classification paradigms, validating the efficacy and advantages of hyperbolic representations for fine-grained sentiment analysis.

0 citationsRead paper

Synchronization-based clustering on the unit hypersphere

Mar 05, 2026

This work addresses the challenging geometric problem of clustering data on the unit hypersphere by introducing, for the first time, a d-dimensional generalized Kuramoto synchronization dynamics model into clustering tasks. By integrating spherical geometric constraints with synchronization mechanisms, the proposed approach effectively captures the intrinsic manifold structure of the data. While maintaining theoretical rigor, the method significantly enhances clustering performance, consistently matching or outperforming state-of-the-art clustering algorithms across multiple synthetic and real-world datasets. These results demonstrate the effectiveness and potential of synchronization dynamics for clustering in non-Euclidean geometric settings.

0 citationsRead paper

UAV-Supported Maritime Search System: Experience from Valun Bay Field Trials

Feb 09, 2026

This work proposes an end-to-end autonomous maritime search system to address the low efficiency and poor robustness of floating object detection in complex oceanic environments. The system uniquely integrates high-fidelity flow field reconstruction driven by computational fluid dynamics with dynamic probabilistic modeling, multi-UAV cooperative search control, and deep learning–based object detection, further enhanced by real-time drifter data for online optimization. Field trials conducted under realistic sea conditions in Valun Bay, Croatia, demonstrated the system’s ability to reliably and efficiently locate floating targets, thereby validating its practicality and robustness in uncertain marine environments.

0 citationsRead paper

Experimental validation of UAV search and detection system in real wilderness environment

Feb 24, 2025

Validating unmanned aerial vehicle (UAV) system performance for search and rescue (SAR) in real wilderness environments remains challenging. Method: This paper proposes an autonomous search framework integrating heat-equation-driven ergodic control (HEDAC), probabilistic search modeling, and YOLOv5-based visual detection, empirically evaluated in a Mediterranean karst wilderness. Contribution/Results: We innovatively couple HEDAC’s ergodic trajectory generation with dynamic target existence probability, jointly optimizing spatial coverage and detection likelihood. Empirical validation with 78 human subjects confirms the efficacy of uniform prior probability modeling. Experiments demonstrate robust, reproducible, and efficient spatial coverage in complex terrain, significantly improving target detection rates. Moreover, YOLOv5 detection outputs exhibit strong correlation with actual search effectiveness, validating the framework’s operational fidelity.

0 citationsRead paper

Recent publications

Latest Papers

Busemann energy-based attention for emotion analysis in Poincaré discs

Apr 08, 2026

Traditional sentiment analysis treats emotions as discrete categories, failing to capture their semantic hierarchy and inherent ambiguity. This work proposes EmBolic, the first end-to-end trainable fully hyperbolic sentiment analysis architecture that integrates hyperbolic geometry with Busemann energy. Operating within the Poincaré disk model, EmBolic employs a Busemann energy–driven attention mechanism to map input text to query points in hyperbolic space and automatically generates boundary key points to represent emotional directions. The model can infer the curvature of the continuous emotion space, achieving high accuracy and strong generalization even with low-dimensional embeddings. Experimental results demonstrate that EmBolic significantly outperforms conventional classification paradigms, validating the efficacy and advantages of hyperbolic representations for fine-grained sentiment analysis.

0 citationsRead paper

Synchronization-based clustering on the unit hypersphere

Mar 05, 2026

This work addresses the challenging geometric problem of clustering data on the unit hypersphere by introducing, for the first time, a d-dimensional generalized Kuramoto synchronization dynamics model into clustering tasks. By integrating spherical geometric constraints with synchronization mechanisms, the proposed approach effectively captures the intrinsic manifold structure of the data. While maintaining theoretical rigor, the method significantly enhances clustering performance, consistently matching or outperforming state-of-the-art clustering algorithms across multiple synthetic and real-world datasets. These results demonstrate the effectiveness and potential of synchronization dynamics for clustering in non-Euclidean geometric settings.

0 citationsRead paper

UAV-Supported Maritime Search System: Experience from Valun Bay Field Trials

Feb 09, 2026

This work proposes an end-to-end autonomous maritime search system to address the low efficiency and poor robustness of floating object detection in complex oceanic environments. The system uniquely integrates high-fidelity flow field reconstruction driven by computational fluid dynamics with dynamic probabilistic modeling, multi-UAV cooperative search control, and deep learning–based object detection, further enhanced by real-time drifter data for online optimization. Field trials conducted under realistic sea conditions in Valun Bay, Croatia, demonstrated the system’s ability to reliably and efficiently locate floating targets, thereby validating its practicality and robustness in uncertain marine environments.

0 citationsRead paper

Experimental validation of UAV search and detection system in real wilderness environment

Feb 24, 2025

Validating unmanned aerial vehicle (UAV) system performance for search and rescue (SAR) in real wilderness environments remains challenging. Method: This paper proposes an autonomous search framework integrating heat-equation-driven ergodic control (HEDAC), probabilistic search modeling, and YOLOv5-based visual detection, empirically evaluated in a Mediterranean karst wilderness. Contribution/Results: We innovatively couple HEDAC’s ergodic trajectory generation with dynamic target existence probability, jointly optimizing spatial coverage and detection likelihood. Empirical validation with 78 human subjects confirms the efficacy of uniform prior probability modeling. Experiments demonstrate robust, reproducible, and efficient spatial coverage in complex terrain, significantly improving target detection rates. Moreover, YOLOv5 detection outputs exhibit strong correlation with actual search effectiveness, validating the framework’s operational fidelity.

0 citationsRead paper