Extended Data Fig. 5: Uniform manifold approximation and projection (UMAP) visualization of RNPU preprocessed dataset.
From: Analogue speech recognition based on physical computing
UMAP is a nonlinear technique for dimensionality reduction that maps high-dimensional data into low-dimensional data while preserving the local structure in data. Where t-SNE uses a Gaussian kernel to measure the similarity of the points in a high-dimensional space, UMAP uses a differentiable kernel, which is a weighted combination of two probability distributions. This figure shows the UMAP visualization of the same data visualized with t-SNE in Fig. 3 of the manuscript.
