Particle Swarm Optimization
Feature Genetic Algorithm (GA) Genetic Programming (GP) Grammatical Evolution (GE) Ant Colony Optimization (ACO)
(PSO)
Variable-sized syntax trees Variable-length linear binary strings A construction graph where vertices or Particles with real-valued positions
Fixed-length linear chromosomes
Representation representing executable computer consisting of 8-bit codons mapped to edges represent discrete solution and velocities moving through a
(e.g., binary bit strings, real numbers).
programs or math expressions. programs via a grammar. components. continuous search space.
Guided by a fitness function and Guided by velocity and position
Evaluated by fitness; uses subtree Uses pheromone trails
Main Heuristics & genetic operators (selection, Uses single-point crossover, bit updates based on personal best (
crossover, mutation, reproduction, (updating/evaporation) and heuristic
Operators crossover, mutation, elitism) [35, mutation, reproduction, and elitism. pbest) and global best (gbest)
and elitism. values to probabilistically build paths.
history]. experiences.
Assesses the quality of a candidate's Assesses how well a program solves A problem-dependent function
Evaluates the mapped phenotype Evaluates path quality. Example: F (s) =
Fitness Function numerical parameters. Example: the problem. Example: Accuracy or evaluating continuous variables.
program. Example: Accuracy in Q/L, where shorter path lengths (L)
(Examples) Shortest distance in a TSP tour [6, mean absolute error against fitness Example: Minimum error rate for
generating correct PIN numbers. yield higher fitness.
history]. cases. decision variables.
Finding optimal values/parameters; Automating program creation; Combinatorial optimization with Continuous optimization where
Best Suited Evolving complex programs using
discrete or numerical optimization Symbolic Regression, classification, discrete variables, like network routing decision variables fall within a
Problems domain-specific grammatical rules.
like the Traveling Salesman Problem. image detection. or the TSP. continuous range.
Randomly generated valid candidate Randomly generated programs built
Initial Population Randomly generated variable-length A colony of artificial ants, each placed A swarm of particles initialized with
solutions within defined from function/terminal sets (e.g.,
Generation binary strings within user-specified limits. at a randomly selected starting node. random positions and velocities.
ranges/encodings. ramped half-and-half method).
Uses a Context-Free Backus-Naur Form
Function Set, Function Set: Operators (e.g., +, -, *,
(BNF) grammar (with Non-terminals,
Terminal Set & N/A /).<br>Terminal Set: Inputs/constants N/A N/A
Terminals, and Production Rules) to map
Grammar (e.g., x, y, 3.5).
binary codons.
A collection of input/output values
Can use input/output pairs to evaluate
Fitness Cases N/A used to evaluate the program (crucial N/A N/A
the generated phenotype.
for Symbolic Regression).
Maximum number of generations, Maximum number of generations, or
Objective function met, or maximum Maximum number of iterations, or CPU Maximum number of iterations, or
Termination Criteria convergence, or a near-optimal near-optimal program generated
number of generations achieved. time limit met. minimum error achieved.
solution is found. (objective met).
Let me know if you need deeper details on any specific algorithm's mechanics!
Feature Genetic Algorithm (GA) Genetic Programming (GP) Grammatical Evolution (GE) Ant Colony Optimization (ACO)
(PSO)
Variable-sized syntax trees Variable-length linear binary strings A construction graph where vertices or Particles with real-valued positions
Fixed-length linear chromosomes
Representation representing executable computer consisting of 8-bit codons mapped to edges represent discrete solution and velocities moving through a
(e.g., binary bit strings, real numbers).
programs or math expressions. programs via a grammar. components. continuous search space.
Guided by a fitness function and Guided by velocity and position
Evaluated by fitness; uses subtree Uses pheromone trails
Main Heuristics & genetic operators (selection, Uses single-point crossover, bit updates based on personal best (
crossover, mutation, reproduction, (updating/evaporation) and heuristic
Operators crossover, mutation, elitism) [35, mutation, reproduction, and elitism. pbest) and global best (gbest)
and elitism. values to probabilistically build paths.
history]. experiences.
Assesses the quality of a candidate's Assesses how well a program solves A problem-dependent function
Evaluates the mapped phenotype Evaluates path quality. Example: F (s) =
Fitness Function numerical parameters. Example: the problem. Example: Accuracy or evaluating continuous variables.
program. Example: Accuracy in Q/L, where shorter path lengths (L)
(Examples) Shortest distance in a TSP tour [6, mean absolute error against fitness Example: Minimum error rate for
generating correct PIN numbers. yield higher fitness.
history]. cases. decision variables.
Finding optimal values/parameters; Automating program creation; Combinatorial optimization with Continuous optimization where
Best Suited Evolving complex programs using
discrete or numerical optimization Symbolic Regression, classification, discrete variables, like network routing decision variables fall within a
Problems domain-specific grammatical rules.
like the Traveling Salesman Problem. image detection. or the TSP. continuous range.
Randomly generated valid candidate Randomly generated programs built
Initial Population Randomly generated variable-length A colony of artificial ants, each placed A swarm of particles initialized with
solutions within defined from function/terminal sets (e.g.,
Generation binary strings within user-specified limits. at a randomly selected starting node. random positions and velocities.
ranges/encodings. ramped half-and-half method).
Uses a Context-Free Backus-Naur Form
Function Set, Function Set: Operators (e.g., +, -, *,
(BNF) grammar (with Non-terminals,
Terminal Set & N/A /).<br>Terminal Set: Inputs/constants N/A N/A
Terminals, and Production Rules) to map
Grammar (e.g., x, y, 3.5).
binary codons.
A collection of input/output values
Can use input/output pairs to evaluate
Fitness Cases N/A used to evaluate the program (crucial N/A N/A
the generated phenotype.
for Symbolic Regression).
Maximum number of generations, Maximum number of generations, or
Objective function met, or maximum Maximum number of iterations, or CPU Maximum number of iterations, or
Termination Criteria convergence, or a near-optimal near-optimal program generated
number of generations achieved. time limit met. minimum error achieved.
solution is found. (objective met).
Let me know if you need deeper details on any specific algorithm's mechanics!
