The Ant Meta-Heuristic Colony Optimization.pptVIP

Seminar Learning Algorithms The Ant Meta-Heuristic Colony Optimization Paper by: Marco Dorigo, Gianni Di Caro Learning and Optimization? Machine learning: learn regularity from given training data, un-/supervised E.g. decision tree, neural networks, clustering, self-organizing map Training is often reduced to optimization E.g. minimize the training error, optimize quantization error, optimize topology, find a minimum model, … Optimization and Learning? Discrete optimization methods like GA, ACO are considered as learning algos Optimize a possibly heuristic or changing function based on simple mechanisms Learn how simple mechanisms affect the function Examples (individuals, ants) are explored during training History of Ant Algorithms Goss et al. 1989, Deneuborg et al. 1990, experiments with Argentine ants Dorigo et al. 1991, applications to shortest path problems Now: established method for various optimization problems Ant Algorithms for shortest path Ants choose paths depending on pheromone After collecting food, paths are marked After some time, the shortest path has the highest probability General Ant Colony Heuristic For discrete optimization problem with: Set of nodes S={1,…,N} Edges for nodes with time-dependent costs States = lists of nodes Neighbourhood of states given by the edges Constraints for feasible states, solutions Costs for solutions given by the edge-costs General Ant Colony Heuristic (II) General Ant Colony Heuristic (III) Ants generation and activity: while resources available: create ant for each ant: initialize let ant run until a solution is found possibly: update pheromone and routing table ACO for shortest path problem Initialize maximum number of ants at source For each ant: Choose next node depending on the pheromone Deposit pheromone on the edge Die at destination Evaporate pheromone ACO for the TSP Randomly place ants at the cities For each ant: Choose a not yet visited city until a tour is completed D