Abstract:Dimensionality reduction techniques are often used by researchers in order to make high dimensional data easier to interpret visually, as data visualization is only possible in low dimensional spaces. Recent research in nonlinear dimensionality reduction introduced many effective algorithms, including t-distributed stochastic neighbor embedding (t-SNE), uniform manifold approximation and projection (UMAP), dimensionality reduction technique based on triplet constraints (TriMAP), and pairwise controlled manifold approximation (PaCMAP), aimed to preserve both the local and global structure of high dimensional data while reducing the dimensionality. The UMAP algorithm has found its application in bioinformatics, genetics, genomics, and has been widely used to improve the accuracy of other machine learning algorithms. In this research, we compare the performance of different fuzzy information discrimination measures used as loss functions in the UMAP algorithm while constructing low dimensional embeddings. In order to achieve this, we derive the gradients of the considered losses analytically and employ the Adam algorithm during the loss function optimization process. From the conducted experimental studies we conclude that the use of either the logarithmic fuzzy cross entropy loss without reduced repulsion or the symmetric logarithmic fuzzy cross entropy loss with sufficiently large neighbor count leads to better global structure preservation of the original multidimensional data when compared to the loss function used in the original UMAP algorithm implementation.

Uniform Manifold Approximation and Projection (UMAP) and its Variants: Tutorial and Survey

UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

Approximate UMAP allows for high-rate online visualization of high-dimensional data streams

Locality Pursuit Embedding

Parametric UMAP embeddings for representation and semi-supervised learning

Deep Manifold Computing and Visualization

Bringing UMAP Closer to the Speed of Light with GPU Acceleration

Deep Manifold Computing and Visualization Using Elastic Locally Isometric Smoothness

ActUp: Analyzing and Consolidating tSNE and UMAP

IsUMap: Manifold Learning and Data Visualization leveraging Vietoris-Rips filtrations

Fuzzy Information Discrimination Measures and Their Application to Low Dimensional Embedding Construction in the UMAP Algorithm

Lens functions for exploring UMAP Projections with Domain Knowledge

On UMAP's true loss function

Relating tSNE and UMAP to Classical Dimensionality Reduction

Dimensionality reduction by UMAP to visualize physical and genetic interactions

Exploring UMAP in hybrid models of entropy-based and representativeness sampling for active learning in biomedical segmentation

Accelerating UMAP for Large-Scale Datasets Through Spectral Coarsening

The Information Geometry of UMAP

Emerging-properties Mapping Using Spatial Embedding Statistics: EMUSES

Inductive Global and Local Manifold Approximation and Projection

HUMAP: Hierarchical Uniform Manifold Approximation and Projection