DataTerminology.comNEW
Also known as: PSO
A population-based stochastic optimization technique inspired by social behaviors like bird flocking and fish schooling.
DESCRIPTION
Particle Swarm Optimization (PSO) maintains a swarm of candidate solutions that move through the search space guided by personal and neighborhood bests. Each particle updates velocity based on inertia, cognitive, and social components. PSO handles nonconvex, nondifferentiable objectives without gradients. It is used for hyperparameter tuning, control, and engineering optimization. Variants include constriction, dynamic topology, and multi-objective PSO. While easy to implement and parallelize, PSO may stagnate or premature converge without parameter tuning and diversity mechanisms.
ALTERNATIVE DEFINITIONS
Wikipedia
Particle swarm optimization (PSO) is a computational method that optimizes a problem by iteratively trying to improve a candidate solution with regard to a given measure of quality.
Emphasizes iterative improvement of candidate solutions.
IEEE
A population-based stochastic optimization technique inspired by social behavior of bird flocking or fish schooling.
Original formulation highlighting biological inspiration.
Scholarpedia
Particle swarm optimization is a stochastic optimization method inspired by social interaction and intelligence.
Stresses social interaction as the guiding principle.
RELATED TERMS
Ant Colony Optimization Differential Evolution Bayesian Optimization Grid Search Stochastic Optimization Swarm Intelligence Fitness Function Hyperparameter Tuning
COMPETING TERMS
Genetic Algorithm Simulated Annealing
RELATIONS
- Broader: Optimization
KNOWLEDGE AREAS
TARGET AUDIENCE
Data Scientist AI Engineer MLOps Engineer
EXTERNAL SOURCES
- Wikipedia - Particle swarm optimization
- IEEE - Seminal PSO paper by Kennedy and Eberhart
- PySwarms - Python library and tutorials for PSO
- MathWorks - PSO algorithm documentation in MATLAB
TOP QUESTIONS
When should I consider Particle Swarm Optimization?
Use PSO for black-box, nonconvex, or nondifferentiable objectives where gradients are unavailable or unreliable.
Category: Application
What are the key PSO hyperparameters?
Inertia weight, cognitive and social coefficients, velocity bounds, and neighborhood topology control exploration and exploitation balance.
Category: Requirements
How does PSO compare to Genetic Algorithms?
PSO uses velocity updates instead of crossover/mutation, often requiring fewer parameters and being easier to implement.
Category: Comparison
What are common PSO pitfalls?
Premature convergence and stagnation occur if diversity is low or parameters reduce exploration too quickly.
Category: Mistakes
Is PSO parallelizable?
Yes. Fitness evaluations for particles can run in parallel, yielding near-linear speedups for expensive objective functions.
Category: Example
How do I detect PSO stagnation?
Monitor lack of improvement in global best and velocity norms; if stuck, reinitialize or increase exploration parameters.
Category: Testing
What is constriction in PSO?
A factor limiting velocity to ensure convergence stability, derived from theoretical analysis.
Category: Concept
Does PSO guarantee a global optimum?
No, it is heuristic and provides no guarantee, though it often finds good solutions in practice.
Category: Risks
Where is PSO used in ML workflows?
Hyperparameter tuning, feature selection, and optimizing non-differentiable metrics are common applications.
Category: Application