Releases: esa/pagmo
1.1.7
1.1.6
Major Changes
This release contains some restructuring of the code and some important added functionality such as the fitness analysis module and the Travelling Salesman Problem support. The release intends to be also a reference for the code base before GSoC and SOCIS 2015.
New utilities
- The module
PyGMO.util.analysisis now available, allowing for fitness landscape analysis
TSP support
- TSPlib support is added so that PaGMO problems can be instantiated from TSPlib xml files
- Static TSPs can be instantiated as PaGMO problems
- Inver_over and nearest neighborhood solvers are available for TSP problems
- Dynamic TSPs and Limited Capacity Vehicle Routing problems are available
- ACO has been removed
New Migration Scheme (BREAKING BACK-COMPATIBILITY)
- The internal representation of a topology has changed (i.e. the boost graph type). Now the graph is a weighted graph with edge weights representing migration probabilities.
- Island objects no longer accept migration probabilities as arguments. Those are now defined in the topology weights.(refer to the new documentation)
Minor Changes
Build system improvments
- The build system has been updated to improve the automated python3 python2 discovery and link correctly libraries to static and dynamic
Travis CI
- Now a basic version of PaGMO is continuosly compiled via TRAVIS CI
V1.1.5
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