Inherent features
- Inherent parallelism
- Stochastic nature
- Adaptivity
- use of feedback
- Autocatalytic in nature
Double bridge experiments: different lengths - the majority ant go though the short path
p1 = (m1+k)^h /(m1+k)^h + (m2+k)^h
p2=1-p1
Basic ideal of ACO
- inspired by the food foraging behavior of ants
- Ants are agents that: - move along between nodes in a graph
- They choose where to go based on pheromone strength (and maybe other things)
- An ant's path represents a specific candidate solution.
- When an ant has finished a solution, pheromone is laid on its path, according to quality of solution.
- This pheromone trail affects behavior of other ants, and this is called 'stigmergy' -a communication method amongst ants
How to implement in program
- Ants: Simple computer agents
- Move ant: Pick next component in the construction of solution
- Trace: Pheromone, delta thor ^k where i,j a global type of infomation
- Memory: mk or tabu k
Genetic ACO algorithm
- Initialize
- repeat // called each iteration
- each ant is positioned on the initial node
- repeat // called each step
- Each ant applied a state transition rule
- Apply the pheromone local update rule // optional
- until all ants have build a complete solution
- apply a local search procedure // optional
- Apply a pheromone global update rule
- until any stopping criterion is met
- Return the best route
- TSP is a metaphor problem for the ant colony
- Is is one of the most studied NP-hard* problems in the combinatorial optimization
- It is easy to explain. So that the algorithm behavior is not obscured by too many technicalities.
*Np-hard are at least as hard as the hardest problems in NP, which can be solved with Non-deterministic Turing machine in Polynomial time.
Ant Colony Optimization (ACO) for TSP
- Each edge is associated a static value based on the edge-cost n(i,j) = 1/Li,j , heuristic information indicating a chance or likeliness to move from node i to j
Well see you again, @MySimpleDiary my note story teller