jMetal project documentation

Author: Antonio J. Nebro <ajnebro@uma.es>

jMetal is an open source Java-based framework for multi-objective optimization with metaheuristics. It includes a wide set of resources, including state-of-the-art multi-objective algorithms, solution encodings, benchmark problems, quality indicators, and utilities for performing experimental studies.

The current stable version is 7.7 (https://github.com/jMetal/jMetal).

jMetal is described in the following papers:

• jMetal: A Java framework for multi-objective optimization. jMetal 4

• Redesigning the jMetal Multi-Objective Optimization Framework. jMetal 5

• Automatic configuration of NSGA-II with jMetal and irace. jMetal 6

• Evolving a Multi-objective Optimization Framework. jMetal 6

Summary of features:

• Multi-objective algorithms: NSGA-II, SPEA2, PAES, PESA-II, OMOPSO, MOCell, AbYSS, MOEA/D, GDE3, IBEA, SMPSO, SMPSOhv, SMS-EMOA, MOEA/D-STM, MOEA/D-DE, MOEA/D-DRA, MOEA/D-D, MOEA/D-IEpsilon, MOCHC, MOMBI, MOMBI-II, NSGA-III, WASF-GA, GWASF-GA, R-NSGA-II, CDG-MOEA, ESPEA, SMSPO/RP, AGEMOEA, AGEMOEA-II, CDG, FAME, MicroFAME, MOSA, DMOPSO, RVEA, RVEA*, iRVEA, RDS-MOEA, random search.

• Single-objective algorithms: genetic algorithm (variants: generational, steady-state), evolution strategy (variants: elitist or mu+lambda, non-elitist or mu, lambda), DE, CMA-ES, PSO (Standard 2007, Standard 2011), Coral reef optimization.

• Parallel models: Synchronous (multi-threaded, Apache Spark), asynchronous

• Variable representations (encodings): binary, real, integer, permutation, mixed

• Problems:

  • Problem families: ZDT, DTLZ (including the DTLZ-Minus variants), WFG, MaF, LSMOP, RE, CRE, RWA, FDA, CEC2009, LZ09, GLT, MOP, LIRCMOP, UF, ZCAT, EBES

  • Real-world constrained problems: the 50 problems of the CEC2021 RWMOP suite (RCM01-RCM50)

  • Classical problems: Kursawe, Fonseca, Schaffer, Viennet2, Viennet3

  • Constrained problems: Srinivas, Tanaka, Osyczka2, Constr_Ex, Golinski, Water, Viennet4, CF1-CF16 (Xiang et al. suite), C-DTLZ, RCM01-RCM50

  • Combinatorial problems: multi-objective TSP, multi-objective knapsack

  • Academic problems: OneMax, OneZeroMax

• Quality indicators: hypervolume, normalized hypervolume, spread, generalized spread, generational distance, inverted generational distance, inverted generational distance plus, additive epsilon, R2, set coverage, error ratio, average Hausdorff distance.

• Support for experimental studies

• Support for automatic algorithm configuration and design (the jmetal-auto sub-project, which is frozen since jMetal 7.6; see Evolver)