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This repository has been archived by the owner on Jul 16, 2024.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

V1.1.4 Release Notes

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

Major Changes

Multi-objective Optimization

The first seven problems of the DTLZ benchmark suite have been implemented and tested. The plotting functions of the population has been improved (it can now be customized for different colors, symbols and sizes). The DTLZ problems have also an additional plot-function which allows a 3d visualization of higher dimensional non-dominated fronts. (only PyGMO) There is also a convergence metric (p-distance) available in order to check how close a population is to the real pareto-optimal front for the ZDT and DTLZ problems.

Algorithms

A new self-adaptive DE algorithm with p-best crossover was added: MDE_pBX.

Problems

The MGA-1DSM problem was changed and allows for two different encodings: tof (time of flight) and alpha (alpha encoding) by now. Moreover, it now works for every central body and can be instantiated also as multi-objective optimization.

Minor Changes

  • A bug in ZDT4 was fixed (problem was just correct for a dimension of 10)
  • A bug causing slow convergence has been removed from the NSGA-II implementation
  • Added options to IPOPT: nlp_scaling_method, mu_init and obj_scaling_factor
  • Added screen output keyword for Monotonic Basin Hopping
  • A new selection policy (selection of random individuals) is now available
  • Evolution time of an island can now be accessed via the method get_evolution_time
  • Removed bugs from computing the crowding distance
  • PaGMO can now compile in debug mode
  • Build system now works on Ubuntu LTS 12.04
  • Removed serialization bug (stochastic problem called for base, not base_stochastic)
  • Added unit tests for serialization
  • Resolved issues with the Boost 1.52 thread library
  • MPI islands made to act more similar to threaded islands by catching more exceptions
Clone this wiki locally