力学与工程科学学院学术报告通知（2025-18）：Tensor feature extraction for image classification through nonnegative Tucker decomposition and high order PCA—暨庆祝河海大学建校110周年学术活动

报告题目：Tensor feature extraction for image classification through nonnegative Tucker decomposition and high order PCA

报告人：Tsung-Lin Lee (台湾中山大学)

报告时间：2025年7月23日10:00开始（北京时间）

报告地点：河海大学江宁校区乐学楼709会议室

主办单位：力学与工程科学学院

欢迎广大师生参加！

报告摘要：

In a classical machine learning, the classification algorithm inputs training data in the form of vectors. For conforming the input format, tensor data are often expanded into high-dimensional vectors. It leads to the loss of spatially related information adjacent to different orders, thus damages the performance of the classification. Several methods have been proposed to extract the feature from tensor data to avoid destroying tensor structure, such as multilinear principal components analysis (MPCA), multilinear discriminant analysis (MLDA), nonnegative Tucker decomposition (NTD) and so on. In this talk, we present a tensor feature extraction method that combines the NTD feature tensor extraction method and high-order PCA algorithm. The experiment of real-world data shows the method improves the classification accuracy and saves the computing time.

智能计算力学与反问题研讨会

2025年7月23日，河海大学江宁校区乐学楼709会议室