Social Optimization


Beyond the Darwinian principles of evolution, there are many other natural phenomena that  could be imitated and used in design of artificial and intelligent systems. Namely the social behavior of both the humans and animals offer a rich source of inspiration. 

We focus on classical swarm intelligence paradigms like Particle swarm optimization or Ant colony optimization and their application in biomedical engineering

Moreover, we adapt the simulation of opinion formation for optimization domain. The artificial agents hold their opinions on various issues that encode  candidate solution of an optimization problem. The agents change their opinions according to simple social rules that give higher importance to fitter neighbors. This leads to emergence of optimization. Such systems have many common elements with cellular automata, neural networks or physical systems (e.g. Ising model).

Example:

 

Social Optimization

Artificial society of 400 agents placed on the 2D grid. Each agent holds opinions on 20 issues that encode a candidate solution of a 20-dimensional optimization problem. The three figures  depict the spatial distribution of fitness value (the lighter the color, the fitter the agent) after 0, 10 and 15 iterations of the social optimizer.


Previous page: Neural Networks
Next page: Downloads