"稀疏优化在机器学习中的若干应用及性能表现"

Abstract In recent years, exploring the sparsity of solutions and other special structures has become a common issue in many computational and engineering areas. Sparsity is much broader than just having few nonzero elements. This paper explores the applications of sparse optimization in machine learning, focusing on fuzzy support vector machine classification models and the proposed fuzzy silhouette index for updating the weights iteratively. The FC, RflikeF, and other performance metrics for ROC indices and identifying target PSM at the same FDR level have shown superior performance compared to mainstream posterior database search methods. This study highlights the significance of non-smooth optimization, sparse optimization, collaborative filtering, Lasso, homotopy algorithms, gene regularization networks, and peptide recognition in machine learning. This research, conducted for a doctoral dissertation at Dalian University of Technology, provides valuable insights into the practical applications of sparse optimization in various machine learning tasks. Keywords: Non-smooth optimization, Sparse optimization, Collaborative filtering, Lasso, Homotopy algorithm, Gene regularization networks, Peptide recognition.