Distribution entropy regularized multimodal subspace support vector data description for anomaly detection

Chuang Wang, Xin Ning, Pengjiang Qian, Wenjun Hu, Jian Yao, Eddie-Yin-Kwee Ng, Khin-Wee Lai, Shitong Wang

Abstract: Highlights•Proposes DEMS-SVDD for multimodal OCC with few abnormal samples.•Introduces two regularization terms for consistency and complementarity.•Introduces entropy-based weights to improve SVDD boundary robustness.•Provides both linear and nonlinear versions of DEMS-SVDD for flexible applications.•Demonstrates superior accuracy of DEMS-SVDD across multiple real-world datasets.

External IDs:doi:10.1016/j.patcog.2025.112478