Topological Characterization of Global Extrema via Persistent Critical Sets

Fuente: Zenodo
Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: SÉRGIO DE ANDRADE, PAULO
Format: Recurso digital
Veröffentlicht: Zenodo 2025
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866901803202772992
author SÉRGIO DE ANDRADE, PAULO
author_facet SÉRGIO DE ANDRADE, PAULO
contents This paper introduces a novel framework for identifying and characterizing global extrema of real-valued functions using tools from topological data analysis, specifically persistent homology. The central challenge in global optimization is distinguishing the global extremum from numerous local extrema, especially in high-dimensional or noisy settings. We propose the concept of "persistent critical sets," which are regions in the function's domain associated with long-lasting topological features in a sublevel or superlevel set filtration. By analyzing the 0-dimensional persistent homology, which tracks the birth and death of connected components, we can isolate features corresponding to significant basins of attraction or peaks. The persistence of these features serves as a robust measure of their significance, allowing for the filtering of topological noise and the identification of regions containing global extrema. We formalize the definition of persistent critical sets and provide an algorithmic approach for their computation. The method's efficacy is demonstrated on several benchmark optimization functions, such as the Rastrigin and Ackley functions, where it successfully isolates the global minimum's basin of attraction even in the presence of significant noise. This gradient-free, multiscale approach offers a new perspective on global optimization, providing not just a single point estimate but a topologically robust characterization of the regions where global extrema are located.
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_17606210
institution Zenodo
language
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle Topological Characterization of Global Extrema via Persistent Critical Sets
SÉRGIO DE ANDRADE, PAULO
This paper introduces a novel framework for identifying and characterizing global extrema of real-valued functions using tools from topological data analysis, specifically persistent homology. The central challenge in global optimization is distinguishing the global extremum from numerous local extrema, especially in high-dimensional or noisy settings. We propose the concept of "persistent critical sets," which are regions in the function's domain associated with long-lasting topological features in a sublevel or superlevel set filtration. By analyzing the 0-dimensional persistent homology, which tracks the birth and death of connected components, we can isolate features corresponding to significant basins of attraction or peaks. The persistence of these features serves as a robust measure of their significance, allowing for the filtering of topological noise and the identification of regions containing global extrema. We formalize the definition of persistent critical sets and provide an algorithmic approach for their computation. The method's efficacy is demonstrated on several benchmark optimization functions, such as the Rastrigin and Ackley functions, where it successfully isolates the global minimum's basin of attraction even in the presence of significant noise. This gradient-free, multiscale approach offers a new perspective on global optimization, providing not just a single point estimate but a topologically robust characterization of the regions where global extrema are located.
title Topological Characterization of Global Extrema via Persistent Critical Sets
url https://doi.org/10.5281/zenodo.17606210