Skip to content
This repository has been archived by the owner on Jul 16, 2024. It is now read-only.

V1.1.5 Release Notes

Jump to bottom
Jacco Geul edited this page Feb 24, 2014 · 1 revision

Major Changes

This release contains many new added functionalities, while mostly keeping compatibility with previous versions of PaGMO. Most new code has been developed during the GSoC 2013, so special thanks go to all the students and mentors (check the Summer of Code 2013 pages).

New Single-objective Optimisation Algorithms

  • sga_gray Simple Genetic Algorithm with gray binary encoding

New Constrained Optimisation Algorithms

  • Constrained Optimisation by Random Evolution (CORE - repair technique)
  • Co-evolution constraints handling technique (adaptive penalty method).
  • Immune system algorithm for constrained optimisation
  • Self-Adaptive Fitness constraints handling technique (adaptive penalty method)

New Multi-objective Optimization

  • pade: Parallel Decomposition (based on the MOEA/D framework)
  • spea2: Strength Pareto Evolutionary Algorithm 2
  • nspso: multiobjective PSO
  • sms_emoa: S-Metric Selection Evolutionary Multiobjective Optimiser Algorithm
  • vega Vector Evaluated Genetic Algorithm

New Problems

The addition of meta problems is a major addition to problems. These are problems that transform other problems

  • Decomposition
  • Shifted
  • Rotated
  • Noisy
  • Robust
  • Constrained to Multi-Objective
  • Constrained to Unconstrained (feasibility, optimality)
  • Constrained to Penalty (death penalty, Kuri penalty, static penalty through penalty coefficients)

Utils

A new module is also added called utils. the module contains some new utilities and will, in the future grow to offer more.

  • Low discrepancy sequences generator (Faure and Halton)
  • Hypervolume indicators (with a variety of algorithms)
  • Population and Algorithm racing

Population class

The population class was enriched with the following methods

  • pagmo::population::repair: repairs an individual using a pagmo::algorithm
  • pagmo::population::race race population individuals using the pagmo::util::racing classes

Back-compatibility

V.1.1.5 keeps back compatibility with previous version except in the construction of the ZDT and DTLZ problems. These are now constructed as follows prob = problem.zdt(prob_id = 1, param_1 = 30) prob = problem.dtlz(prob_id = 1, k = 20, fdim = 4) and NOT: prob = problem.zdt1(30) prob = problem.dtlz1(20,4)

Minor Changes

  • Problem now have a best_x, best_f, best_c properties containing the best known solution. This is optional (the user can decide to implent it).
  • alpha encoding (in problem mga_1dsm_alpha) now samples uniformly at random on the simplex
  • build system now installs headers and static library
  • compilation on windows MSVC 11 works
Clone this wiki locally