PESA: Prioritized experience replay for parallel hybrid evolutionary and swarm algorithms - Application to nuclear fuel

We propose a new approach called PESA (Prioritized replay Evolutionary and Swarm Algorithms) combining prioritized of reinforcement learning with hybrid evolutionary algorithms. hybridizes different swarm algorithms such as particle optimization, evolution strategies, simulated annealing, differential evolution, modular to account for other three by storing their solutions in shared memory, then applying redistribute data between the integral frequent form based on fitness priority values, which significantly enhances sample diversity algorithm exploration. Additionally, greedy is used implicitly improve exploitation close end evolution. features balancing exploration during search parallel computing result an agnostic excellent performance over wide range experiments problems presented this work. also shows very good scalability number processors solving expensive problem optimizing nuclear fuel power plants. PESA's competitive modularity all allow it join family algorithm; unleashing optimization.