-
In this Project I am performing initial research in constructing efficient RL-based DAS meta black box optimizers.
-
It consists of 3 Metaheuristics: Limited-Memory Covariance Matrix Adaptation Evolutionary Strategy, Particle Swarm Optimization and Generalized Generation Gap with Parent-Centric Recombination.
-
In order to simplify transition between them, all 3 algorithms use the same population size.
A (soon to be) Reinforcement Learning Agent is trained to choose the best metaheuristic for current state of optimization.
Much of the code is borrowed from https://github.com/Evolutionary-Intelligence/pypop/tree/main.