Author:
Feng Junning,Liang Yong,Yu Tianwei
Abstract
AbstractDimension reduction is ubiquitous in high dimensional data analysis. Divergent data characteristics have driven the development of various techniques in this field. Although individual techniques can capture specific aspects of data, they often struggle to grasp all the intricate and complex patterns and structures. To address this limitation, we introduceADM (Adaptive graph Diffusion for Metadimension reduction), a novel meta-dimension reduction method grounded in graph diffusion theory. ADM integrates results from diverse dimension reduction techniques to leverage the unique strength of each individual technique. By employing dynamic Markov processes, ADM simulates information propagation for each dimension reduction result, thereby transforming traditional spatial measurements into dynamic diffusion distances. Importantly, ADM incorporates an adaptive mechanism to tailor the time scale of information diffusion according to sample-specific attributes. This improvement facilitates a more thorough exploration of the dataset’s overall structure and allows the heterogeneity among samples.
Publisher
Cold Spring Harbor Laboratory
Reference47 articles.
1. 50 years of data science;Journal of Computational and Graphical Statistics,2017
2. Umap: Uniform manifold approximation and projection for dimension reduction;arXiv preprint,2018
3. How to use t-sne effectively;Distill,2016
4. Understanding how dimension reduction tools work: an empirical approach to deciphering t-sne, umap, trimap, and pacmap for data visualization;Journal of Machine Learning Research,2021
5. Silhouette Analysis for Human Action Recognition Based on Supervised Temporal t-SNE and Incremental Learning