Safety-informed mutations for evolutionary deep reinforcement learning.
Enrico MarchesiniChristopher AmatoPublished in: GECCO Companion (2022)
Keyphrases
- reinforcement learning
- evolutionary process
- mutation rate
- evolutionary computation
- genetic algorithm
- fitness function
- state space
- evolutionary algorithm
- markov decision processes
- function approximation
- model free
- reinforcement learning algorithms
- population size
- evolutionary optimization
- optimal policy
- multi agent
- evolutionary methods
- machine learning
- temporal difference
- optimal control
- action selection
- genetic programming
- supervised learning
- dynamic programming
- learning algorithm
- population dynamics
- learning process
- temporal difference learning
- evolutionary search
- multi agent reinforcement learning
- data sets
- robotic control