Evolutionary algorithms through adaptive decomposition

Program Code: 
1885
Contact: 

Prof Hussein Abbass (h.abbass@adfa.edu.au)

Description of Work: 

Objectives:

Evolutionary algorithms are implicitly parallel search techniques inspired by evolutionary principles. Work in evolutionary algorithms can take many forms including biologically motivated models and statistical modelling of an optimization problem. The objective of this project is to develop decomposition techniques for evolutionary algorithms in optimization and/or machine learning problems. The decomposition needs to be self-organizing, as such, it is not a fixed predefined scheme of decomposition.

Expected Background Knowledge:

  • Knowledge or demonstrated ability to do programming in parallel highperformance computing environment using C or JAVA
  • Good understanding of Optimisation theory

Description of work:

  • A literature review of evolutionary algorithms for optimization and/or machine learning problems
  • A literature review of decomposition techniques
  • Creating a new evolutionary algorithm technique that is based on selforganized decomposition principles
  • Testing the new technique