Dr. W. B. Langdon
Research Information
Bioinformatics
Gene Chip
Zeta
Evolving Kernels
Often complicated behaviour of particle swarm and other population based
stochastic optimisers on difficult problems
can be approximated by a simpler techniques on equivalent landscapes.
Here we have used genetic programming to evolve a low pass filter which is
applied to the real problem to give a smoother equivalent problem.
Trace of PSO swarm (red arrows) and its centre (blue box)
on Rastrigin.
Lower blue curve is plot of equivalent landscape on which a hill climber
(lower blue box) tracks center of the swarm.
More Links
Halting probability video 40 mins
338M
Evolving Benchmarks
Grade-O-Matic
Java native method
universal
essex only