Hierarchical clustering with dot products recovers hidden tree structure

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gray, Annie, Modell, Alexander, Rubin-Delanchy, Patrick, Whiteley, Nick
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917601286815744
author Gray, Annie
Modell, Alexander
Rubin-Delanchy, Patrick
Whiteley, Nick
author_facet Gray, Annie
Modell, Alexander
Rubin-Delanchy, Patrick
Whiteley, Nick
contents In this paper we offer a new perspective on the well established agglomerative clustering algorithm, focusing on recovery of hierarchical structure. We recommend a simple variant of the standard algorithm, in which clusters are merged by maximum average dot product and not, for example, by minimum distance or within-cluster variance. We demonstrate that the tree output by this algorithm provides a bona fide estimate of generative hierarchical structure in data, under a generic probabilistic graphical model. The key technical innovations are to understand how hierarchical information in this model translates into tree geometry which can be recovered from data, and to characterise the benefits of simultaneously growing sample size and data dimension. We demonstrate superior tree recovery performance with real data over existing approaches such as UPGMA, Ward's method, and HDBSCAN.
format Preprint
id arxiv_https___arxiv_org_abs_2305_15022
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Hierarchical clustering with dot products recovers hidden tree structure
Gray, Annie
Modell, Alexander
Rubin-Delanchy, Patrick
Whiteley, Nick
Machine Learning
In this paper we offer a new perspective on the well established agglomerative clustering algorithm, focusing on recovery of hierarchical structure. We recommend a simple variant of the standard algorithm, in which clusters are merged by maximum average dot product and not, for example, by minimum distance or within-cluster variance. We demonstrate that the tree output by this algorithm provides a bona fide estimate of generative hierarchical structure in data, under a generic probabilistic graphical model. The key technical innovations are to understand how hierarchical information in this model translates into tree geometry which can be recovered from data, and to characterise the benefits of simultaneously growing sample size and data dimension. We demonstrate superior tree recovery performance with real data over existing approaches such as UPGMA, Ward's method, and HDBSCAN.
title Hierarchical clustering with dot products recovers hidden tree structure
topic Machine Learning
url https://arxiv.org/abs/2305.15022