Robust and sparse support vector machine via hybrid truncated loss for supervised classification
Researchers have developed a new hybrid truncated loss function for Support Vector Machines (SVMs) to improve classification accuracy and robustness to outliers. This new function, termed $L_{\mathrm{ht}}$, is designed to be both sparse and bounded, addressing limitations of existing convex and non-convex losses. The proposed $L_{\mathrm{ht}}$-SVM model achieves global convergence and reduced computational cost, outperforming other single-view methods in accuracy and noise resistance. The approach has also been extended to multi-view learning as Mv$L_{\mathrm{ht}}$-SVM, showing superior performance in various metrics compared to existing multi-view techniques. AI
IMPACT Introduces a novel loss function for SVMs, potentially improving performance on classification tasks with noisy data.