Salvato in:
Dettagli Bibliografici
Autori principali: Ravuri, Aditya, Lawrence, Neil D.
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2405.17412
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910936131960832
author Ravuri, Aditya
Lawrence, Neil D.
author_facet Ravuri, Aditya
Lawrence, Neil D.
contents This paper shows that dimensionality reduction methods such as UMAP and t-SNE, can be approximately recast as MAP inference methods corresponding to a model introduced in Ravuri et al. (2023), that describes the graph Laplacian (an estimate of the data precision matrix) using a Wishart distribution, with a mean given by a non-linear covariance function evaluated on the latents. This interpretation offers deeper theoretical and semantic insights into such algorithms, and forging a connection to Gaussian process latent variable models by showing that well-known kernels can be used to describe covariances implied by graph Laplacians. We also introduce tools with which similar dimensionality reduction methods can be studied.
format Preprint
id arxiv_https___arxiv_org_abs_2405_17412
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Towards One Model for Classical Dimensionality Reduction: A Probabilistic Perspective on UMAP and t-SNE
Ravuri, Aditya
Lawrence, Neil D.
Machine Learning
Artificial Intelligence
This paper shows that dimensionality reduction methods such as UMAP and t-SNE, can be approximately recast as MAP inference methods corresponding to a model introduced in Ravuri et al. (2023), that describes the graph Laplacian (an estimate of the data precision matrix) using a Wishart distribution, with a mean given by a non-linear covariance function evaluated on the latents. This interpretation offers deeper theoretical and semantic insights into such algorithms, and forging a connection to Gaussian process latent variable models by showing that well-known kernels can be used to describe covariances implied by graph Laplacians. We also introduce tools with which similar dimensionality reduction methods can be studied.
title Towards One Model for Classical Dimensionality Reduction: A Probabilistic Perspective on UMAP and t-SNE
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2405.17412