Deep Reinforcement Learning Based Parameter Control in Differential Evolution

1 code implementation • 20 May 2019 • Mudita Sharma, Alexandros Komninos, Manuel Lopez Ibanez, Dimitar Kazakov

Second, when DDQN is applied as a parameter controller within DE to a different test set of benchmark functions, DDQN uses the trained neural network to predict which mutation strategy should be applied to each parent at each generation according to the DE state.

Q-Learning reinforcement-learning +1