Tucker Decomposition

Tucker decomposition is a tensor factorization method aiming to represent high-dimensional data as a core tensor and a set of factor matrices, effectively reducing dimensionality and revealing underlying structure. Current research emphasizes efficient algorithms for Tucker decomposition, including adaptations for sparse data, noisy observations, and handling of missing values, often incorporating techniques like manifold regularization and proximal gradient methods. These advancements are improving the accuracy and efficiency of applications across diverse fields, such as hyperspectral image fusion, medical image segmentation, and traffic data imputation, by enabling more robust and scalable analysis of complex, multi-dimensional datasets.