Restricted Boltzmann Machine-Assisted Estimation of Distribution Algorithm for Complex Problems
Algorithm 2
Pseudocode of RBMAEDA.
Begin
1. Initialization: at generation , randomly generate candidate solutions to form the initial population
Do while (Termination conditions are not met)
2.Evaluation: According to the real fitness function, calculate the individual fitness of all solutions in , and preserve the best solution in the current population
3.Selection: Select better individuals by using the truncation selection strategy and rank in order to form the dominant subset
4.Modelling: Train a softmax RBM model based on , and then build a probability model based on the softmax RBM according to (8)
5.Surrogate model: Construct the surrogate model based on the softmax RBM. The value is calculated according to (12) to manage the surrogate model, and then the surrogate model estimates the individual fitness according to (10) and (11) to participate in the evolutionary process.
6. Sampling: Produce offspring by sampling the probability model with roulette sampling according to (9)
7. Updating population: Merge the offspring into to form the new population .
