# Illuminating search spaces by mapping elites

**Authors:** Jean-Baptiste Mouret, Jeff Clune  
**arXiv:** [1504.04909](https://arxiv.org/abs/1504.04909)  
**Date:** 2015-04-20  
**Summary page:** https://www.summarizepaper.com/en/paper/1504.04909v1/

## Key points

- Authors Jean-Baptiste Mouret and Jeff Clune introduce the Multi-dimensional Archive of Phenotypic Elites (MAP-Elites) algorithm
- MAP-Elites revolutionizes traditional search algorithms by providing a comprehensive map of high-performing solutions across user-chosen dimensions of variation
- The algorithm illuminates search spaces, allowing researchers to understand how different solution attributes interact to influence performance
- MAP-Elites can explore a larger portion of the search space compared to existing algorithms, leading to the discovery of superior solutions
- The efficacy of MAP-Elites is demonstrated in generating modular neural networks and designing simulated and real soft robots
- The algorithm enhances the identification of optimal solutions by shedding light on the relationship between performance and specific dimensions within solutions

## Summary

In their paper titled "Illuminating search spaces by mapping elites," authors Jean-Baptiste Mouret and Jeff Clune introduce a novel algorithm called Multi-dimensional Archive of Phenotypic Elites (MAP-Elites) that revolutionizes the traditional approach to search algorithms. This groundbreaking algorithm offers a comprehensive map of high-performing solutions distributed across various dimensions of variation chosen by the user, rather than just identifying a single top-performing solution within a search space. By illuminating search spaces, MAP-Elites allows researchers to gain insights into how different attributes of solutions interact to influence performance positively or negatively. For instance, in the pharmaceutical industry, understanding how variations in molecule size and production cost impact drug performance is crucial. One key advantage of MAP-Elites is its ability to explore a larger portion of the search space compared to existing algorithms, resulting in the discovery of superior overall solutions. The authors demonstrate the efficacy of this innovative algorithm across three problem domains: generating modular neural networks and designing both simulated and real soft robots. By shedding light on the relationship between performance and specific dimensions of interest within solutions, providing a set of diverse yet high-performing solutions, and enhancing the identification of optimal solutions, MAP-Elites represents a significant advancement in science and engineering. This pioneering algorithm has the potential to drive progress in various fields by offering a more holistic understanding of complex search spaces and facilitating the discovery of innovative solutions with far-reaching implications.