Soft Computing Swarm Intelligence

							                                  Soft Computing

               http://arf.iyte.edu.tr/~bkumova/teaching/SoftComp

                                     Dr Bora İ Kumova

    İzmir Institute of Technology; Department of Computer Engineering




                               Swarm Intelligence



İYTE; Soft Computing; Spring 2011; Bora İ Kumova                   Page 27/35
                                    Natural Example
• ant colony: probabilistic; shortest path finding
• bird flocking: probabilistic; protection
• animal herding: probabilistic; protection
• fish schooling: probabilistic; protection
• bacteria growth: probabilistic; protection


• co-evolution: competitive
   – predator-prey: survive-die; human-cow
   – parasite-host: survive-survive; flea-human
• co-evolution: co-operative
   – host-symbiont: survive-survive; human-dog
   İYTE; Soft Computing; Spring 2011; Bora İ Kumova   Page 28/35
                             Ant Colony Optimisation

• gene: check point
• chromosome: ant
  – visited check points on route
• reproduction:
  – mutation: none
  – crossover: none
  – communication: more/less pheromone shorter/longer path


• property: dynamic search space; local search
• application: scheduling; vehicle routing

   İYTE; Soft Computing; Spring 2011; Bora İ Kumova    Page 29/35
                         Particle Swarm Optimisation

• gene: check point
• chromosome: particle
  – visited check points on route & best fit value so far
• reproduction:
  – mutation: none
  – crossover: none
  – communication: best fit value among neighbour particles


• property: dynamic search space; global search
• application: multi-objective optimisation; clustering

   İYTE; Soft Computing; Spring 2011; Bora İ Kumova         Page 30/35
                                     Co-Evolution
• co-operative:
   – goals: intersecting
   – objective function: similar
   – population:
        • homogeneous: multi-objective function
• competitive:
   – goals: contradicting
   – objective function: different
   – population:
        • homogeneous: multi-objective function
        • heterogeneous: different objective functions
 İYTE; Soft Computing; Spring 2011; Bora İ Kumova        Page 31/35
                                       Conclusion




• evolution: adaptation in search space
• ant colony: local search
• particle swarm: global search
• co-evolution: decomposition of large search space




 İYTE; Soft Computing; Spring 2011; Bora İ Kumova     Page 32/35