Vector Quantized-Elites: Unsupervised and Problem-Agnostic Quality-Diversity Optimization

Constantinos Tsakonas, Konstantinos Chatzilygeroudis · 2025

Read Vector Quantized-Elites: Unsupervised and Problem-Agnostic Quality-Diversity Optimization by Constantinos Tsakonas, Konstantinos Chatzilygeroudis (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-08057.txt.

Description

Constantinos Tsakonas and Konstantinos Chatzilygeroudis present a 2025 method for quality-diversity search, which seeks many good and varied solutions rather than one best answer. Older methods such as MAP-Elites need a hand-set grid of traits, so users must know the task well in advance. That need limits where the methods can be used.

Their method, called VQ-Elites, learns the trait grid on its own from data. It uses a form of autoencoder to build a neat grid with no task facts given first. Added bounds and shared work help it converge, and two new scores track how much true variety the search finds.

Catalog record

CREATORConstantinos Tsakonas, Konstantinos Chatzilygeroudis
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE55 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-08057

Browse the full library · About the Tower of Babel