Dimensionality Reduction For Data Visualization By T-Sne With Agar Dataset

Abstract

t-SNE (t-distributed Stochastic Neighbor Embedding) is an unsupervised, non-linear dimensionality reduction technique primarily used for visualizing and exploring high-dimensional data. Unlike PCA (Principal Component Analysis), which is a linear technique that focuses on preserving global structure and maximizing variance, t-SNE is designed to preserve local structure — that is, the small pairwise distances or similarities between nearby points in the high-dimensional space.While PCA provides a clear picture of variance and structure in the data, t-SNE gives you a feel or intuition for how data points relate locally. It has become a widely used tool in machine learning and bioinformatics for exploring hidden patterns, such as clusters of similar samples in high-dimensional biological data (e.g., gene expression or microbial morphology).