Dimensionality reduction using elastic measures

With the recent surge in big data analytics for hyperdimensional data, there is a renewed interest dimensionality reduction techniques. In order these methods to improve performance gains and understanding of underlying proper metric needs be identified. This step often overlooked, metrics are typically chosen without consideration geometry data. this paper, we present method incorporating elastic into t-distributed stochastic neighbour embedding (t-SNE) Uniform Manifold Approximation Projection (UMAP). We apply our functional which uniquely characterized by rotations, parameterization scale. If properties ignored, they can lead incorrect analysis poor classification performance. Through method, demonstrate improved on shape identification tasks three benchmark sets (MPEG-7, Car set Plane Thankoor), where achieve 0.77, 0.95 1.00 F1 score, respectively.