TY - GEN
T1 - Adaptive Combination of a Genetic Algorithm and Novelty Search for Deep Neuroevolution
AU - Segal, Eyal
AU - Sipper, Moshe
N1 - Publisher Copyright:
Copyright © 2022 by SCITEPRESS - Science and Technology Publications, Lda.
PY - 2022/1/1
Y1 - 2022/1/1
N2 - Evolutionary Computation (EC) has been shown to be able to quickly train Deep Artificial Neural Networks (DNNs) to solve Reinforcement Learning (RL) problems. While a Genetic Algorithm (GA) is well-suited for exploiting reward functions that are neither deceptive nor sparse, it struggles when the reward function is either of those. To that end, Novelty Search (NS) has been shown to be able to outperform gradient-following optimizers in some cases, while under-performing in others. We propose a new algorithm: Explore-Exploit ?- Adaptive Learner (E2?AL, or EyAL). By preserving a dynamically-sized niche of novelty-seeking agents, the algorithm manages to maintain population diversity, exploiting the reward signal when possible and exploring otherwise. The algorithm combines both the exploitation power of a GA and the exploration power of NS, while maintaining their simplicity and elegance. Our experiments show that EyAL outperforms NS in most scenarios, while being on par with a GA-and in some scenarios it can outperform both. EyAL also allows the substitution of the exploiting component (GA) and the exploring component (NS) with other algorithms, e.g., Evolution Strategy and Surprise Search, thus opening the door for future research.
AB - Evolutionary Computation (EC) has been shown to be able to quickly train Deep Artificial Neural Networks (DNNs) to solve Reinforcement Learning (RL) problems. While a Genetic Algorithm (GA) is well-suited for exploiting reward functions that are neither deceptive nor sparse, it struggles when the reward function is either of those. To that end, Novelty Search (NS) has been shown to be able to outperform gradient-following optimizers in some cases, while under-performing in others. We propose a new algorithm: Explore-Exploit ?- Adaptive Learner (E2?AL, or EyAL). By preserving a dynamically-sized niche of novelty-seeking agents, the algorithm manages to maintain population diversity, exploiting the reward signal when possible and exploring otherwise. The algorithm combines both the exploitation power of a GA and the exploration power of NS, while maintaining their simplicity and elegance. Our experiments show that EyAL outperforms NS in most scenarios, while being on par with a GA-and in some scenarios it can outperform both. EyAL also allows the substitution of the exploiting component (GA) and the exploring component (NS) with other algorithms, e.g., Evolution Strategy and Surprise Search, thus opening the door for future research.
KW - Evolutionary Computation
KW - Genetic Algorithm
KW - Novelty Search
KW - Reinforcement Learning
UR - https://www.scopus.com/pages/publications/85146199048
M3 - Conference contribution
AN - SCOPUS:85146199048
T3 - ICETE International Conference on E-Business and Telecommunication Networks (International Joint Conference on Computational Intelligence)
SP - 143
EP - 150
BT - IJCCI 2022 - Proceedings of the 14th International Joint Conference on Computational Intelligence
A2 - Back, Thomas
A2 - van Stein, Bas
A2 - Wagner, Christian
A2 - Garibaldi, Jonathan
A2 - Lam, H. K.
A2 - Cottrell, Marie
A2 - Doctor, Faiyaz
A2 - Filipe, Joaquim
A2 - Warwick, Kevin
A2 - Kacprzyk, Janusz
PB - Science and Technology Publications, Lda
T2 - 14th International Joint Conference on Computational Intelligence, IJCCI 2022
Y2 - 24 October 2022 through 26 October 2022
ER -