Research of Incremental Dimensionality Reduction Based on Tensor Decomposition Algorithm

For mass or temporal data, it is too large and even impossible for the calculated amount of dimension reduction all at once. Based on text feature graph clusters, first, each text feature graph serves as a second-order tensor. Then, two or more text feature graphs were made up to form a third-order tensor. Moreover, tensor Tucker decomposition is used to study the incremental dimensionality reduction methods of text feature graphs. Finally, experiments on real data sets show that this method is simple and effective for dimensionality reduction of text feature graphs.