MAP-Elites

tags
Quality diversity, Reinforcement learning
papers
(Mouret and Clune 2015; Cully et al. 2015)

MAP-Elites are an example of QD algorithm. The behavior space is discretized in cells and during exploration, only the best “elite” for each cell is kept.

Individuals are added to the grid if they:

  • fill an empty space
  • are better than an existing elite

Bibliography

Cully, Antoine, Jeff Clune, Danesh Tarapore, and Jean-Baptiste Mouret. 2015. “Robots That Can Adapt Like Animals.” Nature 521 (7553). Nature Publishing Group:503–7.

Mouret, Jean-Baptiste, and Jeff Clune. 2015. “Illuminating Search Spaces by Mapping Elites.” arXiv:1504.04909 [Cs, Q-Bio], April.


← Back to Notes