Swarm Intelligence I: Particle Swarm Optimization
Particle Swarm Optimization
PSO Concept
Particle swarm optimization is a swarm intelligence method where each particle moves through the search space using its own experience and information shared by the swarm. It is especially useful for continuous optimization.
PSO State Variables
Update Equations
A common PSO update is and then , where is inertia, and are learning factors, and are random numbers in .
PSO Pseudocode
-
1
Step 1: Initialize particle positions and velocities.
-
2
Step 2: Evaluate particle fitness.
-
3
Step 3: Update each particle's personal best.
-
4
Step 4: Update the swarm's global or neighborhood best.
-
5
Step 5: Update velocity using inertia, cognitive, and social terms.
-
6
Step 6: Update positions and repeat.
PSO Tuning Roles
Inertia weight $w$
- Encourages momentum
- Higher values support exploration
- Lower values support exploitation
Cognitive term
- Pulls particles toward personal best
- Encourages individual learning
What does the personal best in PSO store?
Each particle remembers the best location it has visited.
Correct answer: The best position found by that particle
What do the cognitive and social terms influence in PSO?
They balance self-learning and group learning.
Correct answer: How particles move toward personal and swarm best positions.
PSO is most often applied to which kind of problem?
PSO is commonly used on real-valued optimization tasks.
Correct answer: Continuous numerical optimization
Parameter Sensitivity
PSO can converge too quickly or oscillate if inertia and learning factors are poorly tuned.
Neighborhood Variants
Some PSO variants replace the single global best with neighborhood bests, which can preserve diversity and reduce the chance of premature convergence.