You are here

Warning message

Attention! This event has already passed.

Evolutionary Transitions as a Metaphor for Compositional Search : Definition and Evaluation of a new Optimisation Algorithm

Wednesday, 26 April, 2006 - 17:00
Campus: Brussels Humanities, Sciences & Engineering campus
Faculty: Science and Bio-engineering Sciences
auditorium Q.d
Ann Defaweux
phd defence

This dissertation presents an new evolutionary algorithm inspired from the biological
theory on evolutionary transition: the Evolutionary Transition Algorithm (ETA). This
algorithm uses the biological metaphor of evolutionary transitions to develop a
compositional search evolutionary algorithm that can address optimization problems.
The idea behind compositional search is to evolve parts of the solutions that, through the
interaction with one another, lead toward a full solution for a given problem. To achieve
this goal, it uses the concepts of symbiotic relation described in the biological theory on
evolutionary transitions.

We illustrate the resulting ETA on different test cases. These tests allow to derive
important properties required by a problem to be addressed by the actual ETA. Among
those, the most important one concerns the notion of structured problem that suppose
that certain variables are more correlated than others. Although the experiments
performed here serve the purpose of a proof of concept. The results obtained strongly
suggest that ETA score remarkably well on problem that exibit such structures. These
experiments also allowed us to compare the current algorithm with other evolutionary
algorithm techniques and in particular with another important compistional search
evolutionary algorithm: the Symbiogenetic Model (SEAM). For this latter, we illustrated
that the ETA performed in a similar way on one particular problem but significantly better
on a more general problem.