Thursday, 25 November 2010
Ants gone wild!
soft computing course, and it's terrific! I always loved stuff like genetic algorithms, neural networks and such, but I wasn't prepared for particle swarms: it's been a shocking revelation. I really fell in love with them, I adore the wide spectrum of emerging behaviours those bunch of coordinates exhibit and I feel overwhelmed when I think about creative uses of the model.
Because of the nature of soft computing's solution spaces, it came natural to me to develop everything in C++, by using templates and wild meta-programming. Now I have a set of absolutely general algorithms and in particular an extended PSO which automatically adapts to any kind of solution and relative function to optimize. Everything seemlessly fit in my PLT graphic framework, and let me visually check its goodness with traditional benchmark functions: Rastrigin's and Shaffer's F6. Swarm perfectly finds a way to the optimum in a very natural-looking way.