Unsupervised learning

Miquel Noguer i Alonso, Daniel Bloch and David Pacheco Aznar

6.1 UNIFORM MANIFOLD APPROXIMATION AND PROJECTION FOR DIMENSION REDUCTION

The UMAP is an algorithm for dimensionality reduction. It was first proposed by McInnes et al (2018) with the objective of reducing the dimensionality of the given data. That is, the algo maps data in a way that tries to fit data from one dimension to another dimension, such that ℝnd, with dimensions d < n. The main difference from other dimensionality reduction algorithms is that instead of trying to preserve global or local distances, the UMAP aims to preserve topological distances using fuzzy topology.

The way the algorithm preserves topological characteristics is by first gathering information about the data structure and then using it to modify the data in such a way that the final structure reproduces the original one.

There are three axioms that are assumed to be true (McInnes et al 2018, p. 13).

  1. There exists a manifold on which the data will be uniformly distributed.

  2. The underlying manifold of interest is locally connected.

  3. The primary goal is to preserve a topological structure of the manifold.

The algorithm has three main phases.

  1. Construct a weighted graph depending on the distributional of the original

Sorry, our subscription options are not loading right now

Please try again later. Get in touch with our customer services team if this issue persists.

New to Risk.net? View our subscription options

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here