A1284
Title: t-distributed stochastic neighborhood embedding of tensor data with two applications
Authors: Soohyun Ahn - Ajou University (Korea, South) [presenting]
Abstract: The visualization of high-dimensional data is an essential challenge in numerous fields, and there exists a diverse range of techniques to tackle it. A novel visualization algorithm named matrix t-SNE, addresses the problem of jointly visualizing the rows and columns of matrix-variate data, capturing both row and column features at the same time. The method uses a joint embedding technique that updates both low-dimensional embeddings simultaneously and identifies the nested structure within a particular feature. This is achieved by defining and optimizing a unified loss function that yields a new embedding technique for joint visualization of high-dimensional matrix-variate data in a scatter plot. The proposed algorithm is demonstrated using two real data examples: exergame data and gene expression data.
