Swarm Intelligence II: Ant Colony Optimization

Manual: General · Subject: Evolutionary Programming

Study how virtual ants build solutions using pheromone trails and probabilistic decision-making.

Ant Colony Optimization

ACO Intuition

Ant colony optimization is inspired by the way ants collectively discover short paths using pheromone trails. Artificial ants construct solutions step by step and reinforce promising components with pheromone updates.

ACO Concepts

Pheromone

A learned desirability signal

Heuristic information

Problem-specific attractiveness measure

Trail evaporation

Gradual forgetting of old information

Ant

A constructive search agent

ACO Pseudocode

1

Step 1: Initialize pheromone levels.
2

Step 2: For each ant, construct a solution probabilistically.
3

Step 3: Evaluate solution quality.
4

Step 4: Update pheromones on good solution components.
5

Step 5: Evaporate pheromones to avoid stagnation.
6

Step 6: Repeat.

ACO vs PSO

ACO

Constructs solutions piece by piece
Uses pheromone memory
Often strong for routing and combinatorial problems

PSO

Moves existing candidate positions
Uses velocity and best positions
Often strong for continuous optimization

What is pheromone evaporation used for in ACO?

To erase all memory instantly To prevent overcommitment and stagnation To make solutions invalid To remove heuristics

What does an ant do in ACO?

Which type of problem is ACO especially well suited for?

Routing and combinatorial optimization Only differentiable calculus problems Image-only classification Floating-point matrix inversion

💡

Design Insight

Good heuristic information can dramatically improve ACO performance because it guides ants before pheromone learning becomes strong.

Probability of Choosing an Edge

A common ACO choice rule assigns higher probability to edges with stronger pheromone and better heuristic value, often proportional to $\tau_{ij}^{\alpha}\eta_{ij}^{\beta}$ .

Neuroanatomy Explorer

Memory Deck

Match the Pairs

Concept Constellation