Skip to content
Back to search results

Arithmetic Optimization Algorithm

Introduced by
Laith Abualigah
Year
2020
Type
Mathematical metaheuristic algorithm
Family
Mathematical Optimization
Font size

What does it do?

The Arithmetic Optimization Algorithm (AOA) is a metaheuristic optimization approach introduced by Abualigah et al. in 2020 that leverages mathematical operators (multiplication, division, addition, subtraction) as the inspiration for search strategies.

Unlike nature-inspired algorithms, AOA uses the inherent properties of arithmetic operations to balance exploration and exploitation, making it particularly effective for mathematical optimization problems.

The intuition behind the method

The core insight of AOA is treating optimization as a mathematical process where arithmetic operators drive the search. Multiplication and division operators spread solutions broadly across the search space (exploration), while addition and subtraction operators refine solutions locally (exploitation). This mathematical abstraction eliminates the need for biological metaphors while maintaining effectiveness. The algorithm probabilistically selects between these operator categories based on iteration progress, naturally transitioning from exploration to exploitation.

History

Introduced by Laith Abualigah and collaborators in 2020, AOA was published in Applied Mathematics and Computation. The algorithm represents a departure from purely nature-inspired metaheuristics, instead grounding optimization theory in mathematical operations. This approach provides a fresh perspective on algorithm design and has proven effective across numerous benchmark problems and practical applications.

When to use it

Apply AOA to both continuous and discrete optimization problems, particularly when mathematical properties of the problem are well-understood. Effective for engineering design, machine learning hyperparameter tuning, and combinatorial problems. Preferred when computational simplicity and rapid convergence are priorities.

Limitations and considerations

  • Mathematical abstraction may be less intuitive than nature-inspired approaches for some practitioners
  • Limited theoretical analysis of convergence properties compared to classical optimization methods
  • Performance on high-dimensional problems shows variable results across different problem landscapes

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Simple mathematical foundation without requiring biological inspiration or complex behavioral modeling
  • Mathematical operations are computationally efficient, reducing per-iteration overhead
  • Strong balance between exploration and exploitation through operator selection probabilities
  • Demonstrates competitive performance on diverse benchmark functions and real-world problems

Applications

  • Structural engineering: optimizing beam dimensions, cross-sections, and material distributions
  • Machine learning: tuning neural network weights and bias terms for improved convergence and accuracy
  • Scheduling problems: determining optimal task scheduling and resource allocation in manufacturing
  • Function approximation: finding optimal parameters for basis function expansions and interpolation
  • Electrical grid optimization: minimizing power losses and optimizing generation unit commitment
  • Portfolio management: allocating investments across assets to maximize risk-adjusted returns

How it works

How it works

1

Create an initial population of candidate solutions uniformly distributed in the search space.

2

Apply multiplication and division operators to generate diverse new solutions and explore unexplored regions.

3

Apply addition and subtraction operators to refine solutions around the best discovered position.

4

Evaluate fitness of new solutions and update the population and best solution tracker.

Common pitfalls

  • Operator probability settings that are too aggressive toward multiplication/division can prevent convergence
  • Insufficient exploration early in optimization leads to getting stuck in local optima
  • Inappropriate scaling of decision variables relative to search space bounds affects search efficiency

Frequently asked questions

Why are multiplication and division used for exploration in AOA?

Multiplication and division operations produce larger magnitudes and wider dispersions of values, causing solutions to spread across the search space. This natural divergence supports exploration by preventing premature concentration around a single region.

How does AOA differ from genetic algorithms?

While GA uses selection, crossover, and mutation inspired by biology, AOA uses arithmetic operators directly as search mechanisms. AOA typically converges faster with fewer hyperparameters, though GA may explore more effectively on highly irregular landscapes.

Can AOA handle discrete optimization problems?

Yes, AOA can be adapted for discrete problems through discretization methods such as rounding continuous solutions or mapping to discrete variable indices. However, careful design of the operator application strategy is necessary to maintain algorithm effectiveness.

What is the computational complexity of AOA per iteration?

AOA has linear computational complexity O(N*D) per iteration, where N is population size and D is problem dimensionality. This makes it computationally efficient compared to algorithms requiring pairwise distance calculations.

Related methods

Sources

  1. Abualigah, L., Yousri, D., Abd Elaziz, M., Ewees, A. A., Al-qaness, M. A., & Gandomi, A. H. (2021). Arithmetic optimization algorithm: A new metaheuristic algorithm for solving optimization problems. Applied Mathematics and Computation, 392, 125450.

    Open publication reference
    Copied.