New research explores affinity matrix smoothing for t-SNE neighborhood preservation

A new research paper explores how altering the affinity matrix in t-SNE, a popular dimensionality reduction technique, impacts the preservation of data neighborhoods. The study introduces a parameter, gamma, to smooth or sharpen the probability distribution within the matrix, which is shown to be equivalent to adjusting the Gaussian bandwidth and thus the perplexity. The findings indicate that sharpening the distribution enhances the preservation of very close neighbors, while smoothing improves the representation of broader local neighborhoods, outperforming other multi-scale approaches in certain ranges. AI

RANK_REASON The item is a research paper published on arXiv detailing a new method for t-SNE. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• DagsHub
• Gamma
• Gotit.pub
• Hugging Face
• IArxiv
• Influence Flower
• Litmaps
• ScienceCast
• scite Smart Citations
• t-Distributed Stochastic Neighbor Embedding

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →