Tree-Based Structural Twin Support Tensor Clustering with Square Loss Function

摘要: Most of the real-life applications involving images, videos etc. deals with matrix data (second order tensor space). Tensor based clustering models can be utilized for identifying patterns in as they take advantage structural information multi-dimensional framework and reduce computational overheads well. Despite such numerous advantages, has still remained relatively unexplored research area. In this paper, we propose a novel technique, termed Treebased Structural Least Squares Twin Support Clustering (Tree-SLSTWSTC), that builds cluster model binary tree, where each node comprises proposed Machine (S-LSTWSTM) classifier considers risk minimization alongside symmetrical L2-norm loss function. The approach results time-efficient learning. Initialization on \(k{-}\)means been implemented to overcome instability disseminated by random initialization. To validate efficacy framework, experiments have performed relevant face recognition optical digit datasets.