Solvers
This page summarizes solvers distributed with the SmartCGMS installation.
Meta-Differential evolution (MetaDE)
Based on: https://doi.org/10.1016/j.cmpb.2016.05.011
This solver algorithm is a differential evolution algorithm, which considers meta-parameters (mutation constant, crossover rate, crossover strategy, ...) a part of a genome. Therefore, the algorithm adapts to the data.
Any evolutionary approach relies on a fast and reliable random number generator. The SmartCGMS supports four variants of the MetaDE algorithm, each using one of the following random number generators:
- MT MetaDE - uses C++ standard library implementation of the Mersenne-Twister generator
- Halton MetaDE - uses a custom implementation of Halton sequence generator
- RND MetaDE - uses true random number generator, if available in system (slow, but truly random)
- Xor-Shift* MetaDE - uses the fast XOR-Shift* algorithm to generate random sequences
Particle Swarm Optimization (PSO)
Based on: ISBN 978-1-55860-595-4, https://www.sciencedirect.com/book/9781558605954/swarm-intelligence
This is the original version of the PSO algorithm. The SmartCGMS supports only a Halton sequence generator in conjunction with this algorithm.
Sequential Brute-Force Scan
This algorithm traverses the search space with a fine stepping in each dimension. There is no algoritmic optimization - the algorithm is intended to pose a baseline for all other solvers.
Sequential Convex Scan
This work is yet to be published. Be patient, we will update the page as soon, as the results are public.
RumorOpt
This work is yet to be published. Be patient, we will update the page as soon, as the results are public.