Marine Predators Algorithm in R (mpa)

mpa is an R implementation of the Marine Predators Algorithm (MPA), a nature-inspired metaheuristic optimization algorithm. This package provides a complete implementation of the MPA algorithm as described in the original paper by Faramarzi et al. (2020).

Documentation: https://urbs-dev.github.io/marinepredator/

Installation

You can install the development version of mpa from GitHub:

# Install remotes if not already installed
if (!require("remotes")) {
  install.packages("remotes")
}

# Install marinepredator
remotes::install_github(repo = "urbs-dev/marinepredator")

Overview

The Marine Predators Algorithm (MPA) is inspired by the foraging behavior of marine predators and their interactions with prey. The algorithm follows these key principles:

  1. Lévy and Brownian Movement: Marine predators use Lévy strategy for environments with low prey concentration and Brownian movement for areas with abundant prey.

  2. Velocity Ratios: The algorithm considers different velocity ratios between predators and prey:

    • High velocity ratio (prey moves faster)
    • Unit velocity ratio (similar speeds)
    • Low velocity ratio (predator moves faster)

  3. Memory: Predators remember successful foraging locations and their associates.

  4. Environmental Effects: The algorithm models natural phenomena like eddy formation and FADs (Fish Aggregating Devices).

Usage

Basic Example 

library(marinepredator)

# Optimize the Sphere function (F01)
result <- mpa(
  SearchAgents_no = 25,    # Number of search agents
  Max_iter = 100,          # Maximum iterations
  lb = -100,               # Lower bound
  ub = 100,                # Upper bound
  dim = 30,                # Number of dimensions
  fobj = F01               # Objective function
)

# Print results
print(result)

Using Benchmark Functions

The package includes 23 standard benchmark functions:

# Get function details
details <- get_function_details("F01")  # Sphere function

# Run optimization
result <- mpa(
  SearchAgents_no = 25,
  Max_iter = 100,
  lb = details$lb,
  ub = details$ub,
  dim = details$dim,
  fobj = details$fobj
)

Custom Functions

You can use MPA with your own objective functions:

# Define a custom function
custom_function <- function(x) {
  return(sum((x - c(1, 2, 3))^2))  # Minimum at (1, 2, 3)
}

# Run optimization
result <- mpa(
  SearchAgents_no = 20,
  Max_iter = 50,
  lb = c(-10, -10, -10),
  ub = c(10, 10, 10),
  dim = 3,
  fobj = custom_function
)

Available Functions

Main Algorithm

Helper Functions

Benchmark Functions (F01-F23)

  • F01(): Sphere function
  • F02(): Sum of absolute values and products
  • F03(): Sum of squared terms
  • F04(): Maximum absolute value
  • F05(): Rosenbrock function
  • F06(): Shifted sphere function
  • F07(): Quartic function with noise
  • F08(): Schwefel function
  • F09(): Rastrigin function
  • F10(): Ackley function
  • F11(): Griewank function
  • F12(): Penalized function 1
  • F13(): Penalized function 2
  • F14(): Shekel’s Foxholes
  • F15(): Kowalik function
  • F16(): Six-Hump Camel Back
  • F17(): Branin function
  • F18(): Goldstein-Price function
  • F19(): Hartman 3D function
  • F20(): Hartman 6D function
  • F21(): Shekel 5 function
  • F22(): Shekel 7 function
  • F23(): Shekel 10 function

Demo

See the demo/ directory for complete usage examples:

  • mpaBenchmark.R: Comprehensive examples of using MPA with benchmark functions

Algorithm Parameters

The MPA algorithm has several key parameters:

  • SearchAgents_no: Number of search agents (predators) in the population
  • Max_iter: Maximum number of iterations
  • FADs: Fish Aggregating Devices parameter (default: 0.2)
  • P: Prey movement probability (default: 0.5)

Performance

The algorithm’s performance can be evaluated using the convergence curve:

# Plot convergence curve
plot(result$Convergence_curve, type = "l", col = "blue", lwd = 2,
     main = "MPA Convergence Curve",
     xlab = "Iteration", ylab = "Best Fitness")
grid()

References

Faramarzi, A., Heidarinejad, M., Mirjalili, S., & Gandomi, A. H. (2020). Marine Predators Algorithm: A Nature-inspired Metaheuristic. Expert Systems with Applications, 113377. DOI: 10.1016/j.eswa.2020.113377

License

This package is licensed under the MIT License.

Acknowledgments

Original MATLAB implementation by Afshin Faramarzi and Seyedali Mirjalili

Note that I have use generative AI to produce the initial translation of the MATLAB code into R. For this, I have used Vibe from Mistral AI (Devstral). In a second phase, I have checked the code manually and produced unit tests. In a third phase, I have use Claude code AI to reinforce documentation, especially for a general presentation of test functions.

Support

For issues, questions, or contributions, please open an issue on the GitHub repository.

Citation

If you use this package in your research, please cite the original paper:

@article{faramarzi2020marine,
  title={Marine Predators Algorithm: A Nature-inspired Metaheuristic},
  author={Faramarzi, Afshin and Heidarinejad, Mohammad and Mirjalili, Seyedali and Gandomi, Amir H},
  journal={Expert Systems with Applications},
  volume={113377},
  year={2020},
  publisher={Elsevier}
}