Algorithm guides

One guide per base-level algorithm of Evolver: where it comes from, the variants Evolver offers, its parameter space and default configurations, how to run it, and, backed by a reproducible experiment, where it works well and where it works poorly. The tutorials teach how to use Evolver; these guides teach the algorithms.

Every guide follows the same outline:

1. Idea and origin: the paper and the central mechanism.

2. Variants in Evolver: which parameter selects them.

3. Parameter space: what is specific to the algorithm, and the number of parameters.

4. Default configurations: the files of defaultConfigurations.

5. Running it: from Java (the guide’s class) and from the command line.

6. Where it works well and 7. where it works poorly: an experiment with several runs, with quality indicators, statistical tests and fronts, against a reference algorithm.

7. Tuning notes: which parameters matter most.

8. References.

The code of each guide is a class in package org.uma.evolver.example.algorithms, checked by an integration test.

Algorithm	Encodings	Guide
NSGA-II	Double, Binary, Permutation	NSGA-II
NSGA-III	Double	planned
MOEA/D	Double, Binary, Permutation	planned
SMS-EMOA	Double, Binary, Permutation	planned
MOPSO	Double	planned
RDEMOEA	Double, Permutation	planned
RVEA	Double	RVEA, RVEA* and iRVEA
AGE-MOEA	Double	planned
SSMOEA	Double	planned; see SSMOEA — Steady-State MOEA meanwhile
PAES	Double, Binary, Permutation	planned; see PAES — Pareto Archived Evolution Strategy meanwhile