An optimum solution could be reached from a pool of solutions using genetic algorithms.
The proposed algorithm evolves a pool of solutions using principles of scatter search, interlinked with an iterated local search as an improvement method.
Although current computational docking protocols are able to generate near-native solutions in a reasonable time, the problem of identifying near-native conformations from a pool of solutions remains very challenging.
These approaches are based on the idea of maintaining a pool of solutions during execution.
The main component of MPA is a pool of solutions which serves as long term memory storing all previously found solutions to the problem.
The main idea behind the two approaches is to use a pool of solutions as long-term memory that stores each unique solution found during the search process (Line 17/19 in Algorithm 8/9).
The concept of maintaining a pool of solutions in MSA turns out to be unsuitable for this problem setting, except for very small instances and medium degrees of dynamism.
The reason for the relative disadvantage of MSA compared with dynamic S-VNS reflects the computational demands of evaluating a pool of solutions based on sampled accidents, which is far greater than the demand associated with comparing a single candidate solution against the current incumbent.
The PCR methods, by favoring nondominated solutions and employing population-based evolutionary approaches (such as a genetic algorithm, particle swarm optimization, and ant colony optimization), generate and operate on a pool of candidate solutions, and therefore are capable of identifying multiple Pareto optimal candidate solutions.
For example, given a pool of candidate solutions X of size M, X = { x1, x2,…, x M } T, a candidate solution x i, that is, an individual, can be represented as a binary string x i = [0,0, 1,0,…, 1].
