Fact-checked by Grok 2 weeks ago

Firefly algorithm

The Firefly algorithm (FA) is a nature-inspired optimization technique that simulates the flashing and attraction behavior of fireflies to solve complex optimization problems. Developed by Xin-She and first presented in , it belongs to the class of algorithms and is particularly effective for global search in continuous, non-linear, and high-dimensional spaces. Unlike traditional gradient-based methods, FA relies on population-based exploration, making it robust to local optima traps and suitable for engineering design, image processing, and scheduling tasks. The algorithm operates under three idealized rules derived from firefly biology: all fireflies form a population and are attracted to one another; the perceived of a firefly decreases with distance, with brighter fireflies drawing others toward them; and the objective function value determines each firefly's , where higher corresponds to better solutions for maximization problems (or inverted for minimization). Since its introduction, has been widely adopted and hybridized with other techniques to address real-world challenges, including structural optimization, in , and routing, due to its simplicity, fewer , and parallelizable . Variants such as and have extended its applicability to and constrained problems, while ongoing research—as of 2025—focuses on advanced hybrids like adaptive and multi-modal versions for applications in task scheduling, battery life , and optimization, alongside and to improve in large-scale applications.

Background and Inspiration

Overview and History

The (FA) is a designed for solving problems, which mimics the flashing behavior of fireflies to simulate attraction between individuals based on their relative brightness corresponding to solution quality. Developed by Xin-She Yang while at the , the algorithm was first formulated in and formally introduced in a publication the following year. FA evolved from earlier bio-inspired optimization techniques, such as (PSO), but distinguishes itself through a pairwise mechanism among all agents, contrasting with PSO's velocity-based updates that rely on personal and global best positions. This approach enables more flexible exploration of the search space by allowing every to potentially attract others, promoting efficient in landscapes. Initial evaluations of FA focused on standard benchmark functions, including the Rosenbrock, Michalewicz’s, and Ackley functions, where it demonstrated competitive performance against PSO and other metaheuristics in terms of accuracy and speed. By 2010, the algorithm saw early adoption in engineering optimization tasks, such as structural design problems involving pressure vessels, highlighting its practical utility beyond theoretical testing.

Biological Basis

Fireflies belong to the family Lampyridae within the order Coleoptera and are renowned for their bioluminescent displays, which serve primarily as signals to attract potential mates. Adult fireflies, particularly males, emit flashes from specialized light organs in their abdomens through a involving and , producing in the yellow-green spectrum. The intensity and pattern of these flashes convey information about the emitter's , such as and genetic quality, influencing female ; brighter or more vigorous flashes typically elicit stronger responses from receptive females, who reply with their own flashes to guide males toward them. Flashing patterns exhibit considerable variation across , acting as species-specific codes to prevent hybridization and facilitate recognition during . In many North American species of the genus Photinus, males produce single or double pulses while flying in low, undulating paths, with females responding from perches. In contrast, certain Southeast Asian species, such as Pteroptyx malaccae, demonstrate synchronous flashing where thousands of males in trees coordinate their pulses rhythmically, often with periods around 0.5 seconds, creating wave-like displays that enhance group visibility and potentially amplify individual attractiveness within the population. This collective influences broader interaction dynamics, allowing for efficient location in dense aggregations. The effectiveness of these signals is inherently limited by distance, as bioluminescent light intensity follows the of propagation and undergoes attenuation through and by atmospheric molecules, , and , restricting visibility to tens of meters under typical conditions. Field observations indicate that firefly flashes are reliably detectable up to 20-25 meters, beyond which the signal fades, promoting localized attraction where nearby individuals interact preferentially and reducing interference from distant competitors. Firefly movement incorporates elements, particularly during and initial mate searches, where individuals employ random walks or exploratory flights to cover potential areas in the absence of guiding signals. Larvae, which are predatory and actively snails or other soft-bodied prey, often or swim in irregular patterns influenced by environmental cues but with inherent to discover hidden resources. Adult males similarly exhibit undirected, randomized flight trajectories when no brighter flashes are perceived, enabling broad exploration of habitats before homing in on attractive cues, a that balances local exploitation with global search.

Mathematical Foundations

Light Intensity and Attractiveness

In the Firefly Algorithm, serves as a metaphorical of a solution's quality, decreasing with from the source to model how fainter signals diminish over . The I at a distance r from a firefly with initial I_0 is given by the formula I(r, I_0) = I_0 e^{-\gamma r^2}, where \gamma > 0 is the light absorption coefficient that governs the rate of decay, and r denotes the Cartesian between two fireflies i and j in the search space. This Gaussian-like decay reflects an idealized approximation of light in a medium, ensuring that nearby fireflies perceive stronger signals while distant ones see negligible influence. The perceived light intensity of each firefly is directly tied to the objective function value f(\mathbf{x}) of its corresponding solution position \mathbf{x}, such that brighter fireflies represent superior (higher f for maximization problems). Specifically, the is idealized as proportional to f(\mathbf{x}), meaning fireflies with better values emit stronger "" and thus attract others more effectively. This linkage transforms the optimization landscape into a dynamic attraction field, where solution improvement drives increased visibility and influence within the swarm. Attractiveness, which dictates the strength of mutual between fireflies, follows a similar distance-dependent and is defined as \beta(r) = \beta_0 e^{-\gamma r^2}, with \beta_0 representing the maximum attractiveness when r = 0 (i.e., at ). Attractiveness models the of influence with distance in a manner analogous to . The shared \gamma in both formulas unifies the models, allowing the to modulate the algorithm's -exploitation —higher values of \gamma cause rapid , favoring localized searches around good solutions, while lower values enable broader global by sustaining over larger distances.

Position Update Mechanism

In the Firefly Algorithm (FA), the position update mechanism governs how each firefly i adjusts its position in the search space based on interactions with other fireflies, simulating the attraction of fireflies to brighter counterparts. For every pair of fireflies i and j, firefly i evaluates the relative attractiveness of j; if j is brighter (i.e., has a superior objective function value), i moves toward j. This pairwise comparison ensures that movement is directed toward potentially better solutions, promoting exploitation of promising regions. If no firefly is brighter than i, it performs a to facilitate exploration. The core of the position update is captured by the equation: \mathbf{x}_i^{t+1} = \mathbf{x}_i^t + \beta (\mathbf{x}_j^t - \mathbf{x}_i^t) + \alpha \boldsymbol{\epsilon}_i^t where \mathbf{x}_i^t denotes the position vector of firefly i at iteration t, \beta represents the attractiveness (computed based on distance), and \alpha \boldsymbol{\epsilon}_i^t is the randomization component with \alpha as the randomization strength (typically between 0 and 1) and \boldsymbol{\epsilon}_i^t as a random vector drawn from a uniform distribution in [-0.5, 0.5] or, in enhanced variants, a Lévy distribution to enable larger jumps for better global search. The attraction term \beta (\mathbf{x}_j^t - \mathbf{x}_i^t) drives convergence toward superior positions, while the stochastic term introduces variability to avoid local optima. This update is applied sequentially for all brighter j, with the final position reflecting the cumulative effect. To maintain feasibility within the defined search space bounds, boundary handling is applied post-update. Common strategies include , where positions exceeding upper or lower limits are mirrored back (e.g., if x > b_u, then x' = 2b_u - x), or wrapping via operation to fold the position into the feasible interval. These methods prevent divergence and ensure solutions remain valid, with often preferred for preserving toward . The parameter \alpha critically balances exploitation and exploration: higher values emphasize random perturbations for broad searching in early iterations, while decreasing \alpha over time (e.g., linearly from 1 to 0) shifts focus to fine-tuning via attraction, aiding convergence. This adaptive tuning enhances the algorithm's efficiency across diverse optimization landscapes.

Algorithm Implementation

Step-by-Step Procedure

The Firefly Algorithm (FA) operates as an iterative optimization process that simulates the flashing behavior of fireflies to explore and exploit the search space for global optima. The procedure begins with the initialization of a population of firefly agents, each representing a potential solution in the D-dimensional problem space, and proceeds through repeated cycles of attraction and movement until a predefined stopping condition is met. This step-by-step flow ensures progressive refinement of solutions by leveraging relative attractiveness based on objective function evaluations.

Initialization Phase

The algorithm starts by defining the objective function f(\mathbf{x}) to be optimized, where \mathbf{x} = (x_1, x_2, \dots, x_D)^T represents a D-dimensional . A population of N fireflies is randomly generated within the specified search bounds, typically using for each to ensure diverse initial coverage of the feasible space. For each firefly i (where i = 1, 2, \dots, N), the I_i is computed as I_i = f(\mathbf{x}_i), assuming higher corresponds to brighter intensity for maximization problems (or inverted for minimization). Additionally, the light absorption coefficient \gamma is set, which governs how attractiveness diminishes with distance. This phase establishes the starting configuration without any movement, preparing for iterative updates.

Main Iterative Loop

The core of the algorithm unfolds in a loop that runs for a maximum number of generations, denoted as MaxGeneration. For each :
  1. Evaluate the current light intensities I_i for all fireflies based on their using the objective function.
  2. For every pair of fireflies i and j (with j = 1 to N and i = 1 to N, often implemented to consider j > i for to avoid redundant pairwise computations):
  3. After processing all pairwise attractions, rank the fireflies by their intensities and identify the current global best solution.
This pairwise comparison ensures that less bright fireflies are drawn towards brighter ones, promoting while the prevents premature trapping in local . The loop repeats, progressively improving the population's overall .

Termination Criteria

The iterative process terminates upon reaching the maximum number of generations (MaxGeneration), which serves as the primary stopping criterion to balance and computational efficiency. Alternative or supplementary criteria may include achieving a , such as when the best value stabilizes below a predefined over several iterations, or exhausting a computational like total function evaluations. Upon termination, the algorithm outputs the best firefly position as the approximate global optimum, often followed by post-processing steps such as of history or solution validation.

Pseudocode Representation

The following pseudocode outlines the complete procedure in a structured, flowchart-like manner:
BEGIN
  Define objective function f(x) where x = (x1, ..., xD)^T
  Set parameters: N (number of fireflies), MaxGeneration, γ ([absorption coefficient](/page/Coefficient)), α ([randomization parameter](/page/Parameter)), β0 (attractiveness at r=0)
  Generate [initial](/page/Initial) [population](/page/Population) of N fireflies xi (i=1 to N) randomly in search space
  For i=1 to N:
    Evaluate Ii = f(xi)
  t = 1
  While t <= MaxGeneration:
    For i=1 to N:  // For each firefly
      For j=1 to N:  // Consider all other fireflies (or j>i for efficiency)
        If Ij > Ii:
          Compute distance rij = ||xi - xj||
          Compute attractiveness β = β0 * exp(-γ * rij^2)
          Update position: xi = xi + β*(xj - xi) + α*(rand - 0.5)  // Simplified update with [randomization](/page/Randomization)
          Evaluate new Ii = f(xi)
        End If
      End For
    End For
    Rank fireflies by Ii and find current best
    t = t + 1
  End While
  Postprocess results: Output best xi and f(xi)
END
This representation captures the essential distance-based attractions and iterative ranking, with the position update briefly referenced as per the mechanism's core role in movement.

Key Parameters and Tuning

The Firefly Algorithm (FA) relies on several core parameters that govern its exploration and exploitation behaviors. The number of fireflies, denoted as N, represents the and typically ranges from 10 to 50, with values around 20 often used in tests to balance diversity and computational efficiency. The randomization parameter \alpha, which controls the step size in random movements, is usually set between 0.1 and 1, with an initial value of 0.2 commonly employed; it can be scaled by the problem's dimensionality to prevent excessive wandering. The light absorption coefficient \gamma determines the decay of attractiveness with and is selected from 0.1 to 10 based on the optimization problem's scale, often starting at 1 to ensure moderate convergence. Finally, the initial attractiveness \beta_0 is generally fixed at 1, representing the maximum attraction when fireflies are at zero . Tuning these parameters involves empirical testing on functions and to assess their impact on performance. For instance, studies using methods like sampling, Quasi-Monte Carlo, and across diverse s (e.g., , Rosenbrock) have shown that optimal values for \alpha_0 = 1, \gamma between 0.1 and 1, and \beta = 1 yield consistent results, with no significant differences arising from the tuning approach itself. analyses reveal that larger N enhances diversity but increases computational cost, while \gamma's value critically affects : higher \gamma promotes search, and lower values maintain . Practitioners often perform iterative tests on problem-specific s to fine-tune these, prioritizing \gamma and \alpha for balancing speed and solution quality. Adaptive variants dynamically adjust parameters during execution to mitigate stagnation. A common strategy anneals \alpha over iterations t up to maximum T, such as \alpha(t) = \alpha_0 (1 - t/T), starting from \alpha_0 = 1 and decreasing to encourage initial exploration followed by exploitation. Similarly, \gamma can be adapted based on average distances among fireflies, e.g., \gamma_i = \gamma_0 (d_i / d_{\max})^\beta with \gamma_0 = 1 and \beta = 2, to tailor absorption to the current . These adaptations have demonstrated superior performance on benchmarks compared to static settings. Common pitfalls in parameter selection include excessively high N, which escalates runtime without proportional gains in solution accuracy, and low \gamma, which diminishes attractiveness and risks reducing the algorithm to ineffective random walks. Overly aggressive \alpha without annealing can also lead to oscillatory behavior, underscoring the need for problem-tailored calibration.

Applications

Engineering and Design Optimization

The Firefly algorithm has found significant application in structural optimization within engineering, particularly for antenna design and truss structures, where the goal is often to minimize structural weight while satisfying stress and displacement constraints. In antenna design, the algorithm optimizes parameters such as element spacing and excitation amplitudes to reduce sidelobe levels and achieve desired radiation patterns. For instance, a 2012 study applied the Firefly algorithm to nonuniformly spaced linear antenna arrays, achieving a sidelobe level of -23.5 dB for a 20-element array, which outperformed particle swarm optimization (PSO) in both solution quality and convergence speed, requiring fewer iterations to reach the global optimum. Similarly, in truss structure optimization, an accelerated variant of the Firefly algorithm was developed in 2013 to perform size optimization on benchmark problems like the 10-bar and 72-bar trusses, resulting in lighter designs that met stress limits more efficiently than the standard algorithm, with improved convergence rates due to reduced randomness in firefly movement. In power systems engineering, the Firefly algorithm addresses economic-emission dispatch problems by minimizing a combined of fuel costs and emissions subject to nonlinear constraints such as and generator limits. This approach leverages the algorithm's ability to handle landscapes, leading to faster compared to genetic algorithms () in benchmark tests. A implementation on a 6-unit system demonstrated optimal costs of $603.35/hour for a 1 MW , with computation times under 3 seconds—significantly quicker than GA and PSO equivalents—while considering transmission losses. For scheduling in manufacturing engineering, the Firefly algorithm excels in job-shop optimization, where it manages multi-objective trade-offs between metrics like makespan, machine workload, and resource utilization. A hybrid discrete Firefly algorithm proposed in 2014 for flexible job-shop scheduling incorporated local search enhancements to navigate discrete solution spaces, outperforming genetic algorithms and tabu search on standard benchmarks by balancing critical machine loads and total workloads more effectively under resource constraints. A notable real-world engineering application occurred in 2017 for inverse heat conduction problems, where the Firefly algorithm, combined with the Newton method, estimated boundary conditions in transient heat transfer scenarios. This capability highlights the algorithm's robustness in ill-posed inverse problems common in design and systems.

Other Domains

The Firefly algorithm has been applied to image processing tasks, particularly for multilevel thresholding and segmentation, where it optimizes histogram-based thresholds to enhance quality and separate regions effectively. In one early , the algorithm was used to maximize in images by selecting optimal threshold levels through the flashing attraction mechanism, demonstrating superior performance over traditional methods like in terms of and processing speed on benchmark images. This approach extended to color images by independently optimizing histograms for red, green, and blue channels, enabling efficient segmentation for applications such as analysis. In , the Firefly algorithm facilitates by treating feature subsets as firefly positions and using attractiveness to converge on subsets that minimize classification error while reducing dimensionality. Applied to UCI repository datasets, it has shown improved performance compared to genetic algorithms, with reduced feature sets enhancing computational efficiency. Additionally, the algorithm optimizes training by adjusting weights and biases as position updates, leading to faster convergence and better generalization on tasks like , as evidenced in feed-forward networks trained on datasets. In bioinformatics, the Firefly algorithm aids by modeling the attraction between fireflies to simulate energy minimization in lattice-based folding models, where brighter solutions represent lower-energy conformations. This method approximates native protein folds for sequences of length around 48-61 residues using simplified H-P energy functions, outperforming optimization in search space exploration. The algorithm's population-based search effectively navigates the rugged energy landscape of , providing insights into tertiary structure formation. Post-2020 applications have extended the Firefly algorithm to handling in environments, incorporating variants like class center-based approaches to impute values efficiently in high-dimensional datasets while considering attribute correlations. In forecasting, Firefly variants have been used to optimize support vector parameters for electric load prediction, achieving low mean absolute percentage errors on half-hourly datasets. As of 2025, ongoing research integrates Firefly algorithm with for enhanced optimization in large-scale AI tasks, such as hyperparameter tuning in neural networks.

Variants and Extensions

Discrete and Binary Versions

The binary firefly algorithm (BFA) adapts the continuous firefly algorithm to binary search spaces by representing firefly positions as strings, where each bit indicates the (1) or exclusion (0) of an element, such as features in selection tasks or items in optimization problems. Introduced in by Palit et al. for cryptanalyzing the knapsack , BFA employs transfer functions to map the continuous position updates to decisions, enabling exploration of discrete solution spaces. Common transfer functions include the , defined as S(x) = \frac{1}{1 + e^{-x}}, which converts the continuous movement step into a probability for flipping bits (e.g., a bit changes from 0 to 1 if a uniform random number is less than S(x)), and the V-shaped function, which maps values to the interval [-1, 1] and uses absolute value for updates to encourage exploitation around promising solutions. In discrete applications, the firefly algorithm handles problems like the traveling salesman problem (TSP), where firefly positions represent permutations of cities, updated via operators such as inversion or swap mutations to mimic movement toward brighter fireflies, as proposed by Jati and Suyanto in 2011. For the , positions encode subsets of items as binary vectors, with updates focusing on feasible solutions that respect capacity constraints, demonstrating effectiveness in as shown in early adaptations for . The pairwise distance r between fireflies, originally Euclidean in continuous spaces, is replaced by the (the number of differing bits) in binary versions, adapting the attractiveness function to \beta = \beta_0 e^{-\gamma r^2}, where r counts bit differences to quantify solution dissimilarity and guide attraction toward superior subsets. Performance evaluations in feature selection tasks highlight BFA's advantages, where it often achieves higher classification accuracy with fewer selected features compared to binary particle swarm optimization (BPSO) on various UCI datasets. These gains stem from BFA's multi-directional attraction mechanism, which enhances diversity and convergence in high-dimensional binary spaces over BPSO's velocity-based updates.

Hybrid and Multi-Objective Adaptations

The multi-objective firefly algorithm (MOFA) extends the standard firefly algorithm to handle multiple conflicting objectives by incorporating Pareto dominance concepts. Introduced in 2012 by Xin-She Yang, MOFA replaces the single fitness function with non-dominated sorting to rank fireflies, where attraction is determined based on Pareto fronts rather than a scalar objective value, enabling the generation of diverse non-dominated solutions. This adaptation maintains the core attraction mechanism but archives elite solutions to balance exploration and convergence in multi-objective spaces. Hybrid variants of the firefly algorithm integrate elements from other metaheuristics to enhance performance in specific optimization scenarios. A notable Firefly-PSO hybrid, proposed in 2014, combines the velocity update from with firefly attraction to accelerate convergence while preserving global search capabilities, particularly effective for controller design problems where rapid solution refinement is required. Similarly, a Firefly-GA hybrid from 2014 incorporates crossover operations into the firefly movement to promote population diversity and escape local optima, applied successfully to facility location problems with discrete constraints. These hybrids leverage the firefly's local intensification with complementary mechanisms from PSO and GA for improved robustness. Chaotic firefly algorithms introduce to mitigate premature convergence by replacing the randomization parameter α with chaotic maps, enhancing diversity in the search process. Developed in 2015, a chaotic variant using multi-population strategies and chaotic perturbations for α was applied to dynamic optimization problems, where environments change over time, demonstrating superior tracking of shifting compared to standard firefly implementations. This approach avoids stagnation in local through the ergodic properties of chaotic sequences, making it suitable for time-varying applications like scheduling. MOFA and its hybrids exhibit scalability in high-dimensional multi-objective problems, such as involving trade-offs between conflicting goals like cost minimization and quality maximization. These adaptations highlight the algorithm's flexibility for complex, real-world systems with numerous variables and goals. Recent developments (2020–2025) include variants like a gender-classified firefly algorithm for improved in multimodal problems and hybrids integrating firefly with models, such as ResNet for recognition, enhancing performance in engineering and applications.

Performance Analysis

Comparisons with Other Metaheuristics

The Firefly algorithm (FA) differs from (PSO) in its core mechanics, employing pairwise attraction based on firefly brightness rather than PSO's velocity updates influenced by inertia and personal/global best positions, which can lead to premature convergence in complex landscapes. Empirical studies on benchmark functions such as , Rosenbrock, Rastrigin, Griewank, and Schaffer's f6 demonstrate that FA requires 79-98% fewer generations to achieve convergence compared to PSO, with execution times 23-30% faster in implementations like . For instance, on the Schaffer's f6 function, FA converges in 6-9 generations versus significantly higher for PSO, highlighting FA's superior efficiency in multi-modal optimization. In comparison to genetic algorithms (GA), FA features fewer tunable parameters, avoiding the need for explicit crossover and mutation rates that GA relies on for population diversity, making FA simpler to implement for tasks. However, GA often excels in highly rugged, discontinuous landscapes due to its evolutionary operators, while FA demonstrates faster convergence on unimodal and moderately multi-modal functions. Reviews indicate FA achieves higher success rates and efficiency than GA on multi-modal benchmarks, including CEC test suites and the , with binary variants of FA outperforming GA in applications like for large datasets. On the Ackley function, FA variants show improved convergence accuracy over GA. Relative to ant colony optimization (ACO), which uses pheromone trails for path-based search in discrete problems, FA's light-based attraction mechanism introduces greater randomness, enabling better adaptation to dynamic environments without the stagnation risks associated with pheromone evaporation in ACO. FA typically converges faster with lower time complexity (O(m²) for m fireflies) and fewer parameters, though ACO provides stronger global search in combinatorial tasks like the traveling salesman problem. Some studies report FA advantages over ACO in certain scheduling and path planning applications, with higher success rates in multi-objective benchmarks. Benchmark tests from 2009-2015 CEC suites and classical functions reveal FA's strengths on multi-modal problems like Rastrigin, where variants achieve lower error rates than PSO and , but it can be slower on high-dimensional unimodal functions such as due to exhaustive pairwise evaluations. A 2020 systematic review confirms FA's superiority over baselines on Rastrigin and Ackley, with improvements in exploration via Lévy flights, though high dimensionality amplifies computational costs relative to PSO's linear updates. These results underscore FA's balance of speed and robustness in , as validated across 13-30 dimensional tests. Recent research (2023-2025) continues to explore hybrid variants to enhance performance in high-dimensional spaces.

Strengths and Limitations

The Firefly Algorithm (FA) exhibits several inherent strengths that contribute to its utility in optimization tasks. Its implementation is notably simple, requiring only basic operations such as distance calculations and attractiveness evaluations, which makes it accessible for practitioners without extensive computational resources. Furthermore, the algorithm's population-based nature allows for automatic subdivision into subgroups, enabling effective exploration of multimodal landscapes by simultaneously pursuing multiple optima without explicit niching mechanisms. As a derivative-free metaheuristic, FA is particularly effective for non-convex, continuous optimization problems where gradient information is unavailable or unreliable, often outperforming traditional methods in global search efficiency. Despite these advantages, FA faces significant limitations in practical deployment. The core update mechanism involves pairwise comparisons among all fireflies, resulting in a computational complexity of O(N²) per —where N is the —dominated by and attractiveness computations, which can become prohibitive for large-scale problems. Additionally, the algorithm is highly sensitive to parameter settings, such as the randomization coefficient α, attractiveness β₀, and light absorption γ; suboptimal choices can lead to premature stagnation or excessive , hindering . Theoretical analyses underscore FA's global search capabilities while revealing practical challenges. Early convergence studies provide proofs of the algorithm's ability to achieve global optima under certain conditions, demonstrating its convergence properties through Markov chain-like transitions in the search space. However, empirical evaluations indicate slower in noisy or dynamic environments, where the fixed attractiveness decay can trap solutions in local optima due to insufficient . The original 2009 formulation of FA lacks inherent scalability for high-dimensional or scenarios, as its sequential pairwise operations do not parallelize efficiently without modifications. Variants have incorporated parallelization to address O(N²) bottlenecks, improving efficiency in large-scale applications. Recent studies (up to 2025) highlight ongoing efforts to mitigate issues like through enhanced strategies.

References

  1. [1]
    [1003.1466] Firefly Algorithms for Multimodal Optimization - arXiv
    Mar 7, 2010 · This paper intends to provide a detailed description of a new Firefly Algorithm (FA) for multimodal optimization applications.Missing: original | Show results with:original
  2. [2]
    Firefly Algorithms for Multimodal Optimization - SpringerLink
    This paper intends to provide a detailed description of a new Firefly Algorithm (FA) for multimodal optimization applications.
  3. [3]
    A Systematic Review on Firefly Algorithm: Past, Present, and Future
    Sep 29, 2020 · Firefly algorithm (FA) is one of the popular algorithm of Swarm Intelligence domain that can be used in most of the areas of optimization.
  4. [4]
    (PDF) Firefly Algorithms for Multimodal Optimization - ResearchGate
    Aug 7, 2025 · This paper intends to provide a detailed description of a new Firefly Algorithm (FA) for multimodal optimization applications.
  5. [5]
    Sexual Selection: Signals to Die for - ScienceDirect.com
    Male Photinus fireflies attract mates by flying around in the dark emitting flashes of bioluminescent light, with species-specific flash rates and durations.
  6. [6]
    Determinants of reproductive success across sequential episodes of ...
    Aug 22, 2012 · In Photinus fireflies, females preferentially respond to males based on their bioluminescent courtship signals, but previous work has shown ...
  7. [7]
    Artificial light impacts the mate success of female fireflies - PMC
    Aug 10, 2022 · Adults of most species have since co-opted this bioluminescent ability for courtship signals: glows or patterns of flashes used to attract mates ...
  8. [8]
    Mechanism of rhythmic synchronous flashing of fireflies ... - PubMed
    In Thailand, male Pteroptyx malaccae fireflies, congregated in trees, flash in rhythmic synchrony with a period of about 560 +/- 6 msec (at 28 degrees C).
  9. [9]
    A spiking neuron model for synchronous flashing of fireflies
    It was reported that thousands of fireflies in southeast Asia and the Pacific congregate in trees and produce a synchronous flashing.
  10. [10]
    A model for the collective synchronization of flashing in Photinus ...
    Oct 26, 2022 · Biology of synchronous flashing of fireflies. ... fireflies of southeast Asia may use anticipatory time-measuring in synchronizing their flashing.
  11. [11]
    [PDF] Protocols and guidelines for measuring indices of abundance in ...
    Sampling unit dimensions: maximum detection distance. 20-25 meters. Although flashing fireflies can sometimes be seen and counted from great distances. ( ...
  12. [12]
    The size of signal detection and emission organs in a synchronous ...
    Oct 27, 2020 · The size of signal detection and emission organs in a synchronous firefly ... detection distance of the usually faint glows produced by the ...<|control11|><|separator|>
  13. [13]
    Random walk models in biology | Journal of The Royal Society ...
    Apr 15, 2008 · We review models and results relating to the movement, dispersal and population redistribution of animals and micro-organisms.
  14. [14]
    A comprehensive review and call for studies on firefly larvae - PMC
    Sep 20, 2021 · This paper reviews and indicates key gaps in the biology of firefly larvae based on available literature.Missing: random | Show results with:random
  15. [15]
    Random Search Wired Into Animals May Help Them Hunt
    Jun 11, 2020 · The nervous systems of foraging and predatory animals may prompt them to move along a special kind of random path called a Lévy walk to find food efficiently ...
  16. [16]
    Firefly Algorithm, Lévy Flights and Global Optimization - SpringerLink
    In this paper, we intend to formulate a new metaheuristic algorithm by combining Lévy flights with the search strategy via the Firefly Algorithm.Missing: original | Show results with:original
  17. [17]
    [PDF] BOUNDARY STRATEGIES FOR FIREFLY ALGORITHM ANALYSED ...
    After each step of an algorithm, in this case after position update of a firefly, the new position should be checked if it lies in the appropriate range or ...
  18. [18]
  19. [19]
    Nonuniformly Spaced Linear Antenna Array Design Using Firefly ...
    Apr 22, 2012 · A nonuniformly spaced linear antenna array with broadside radiation characteristics is synthesized using firefly algorithm and particle swarm optimization.
  20. [20]
    A new accelerated firefly algorithm for size optimization of truss ...
    Aug 7, 2025 · An Accelerated Firefly Algorithm (AFA) for fast size optimization of truss structures is proposed in this paper.
  21. [21]
    Application of the Firefly Algorithm for Solving the Economic ...
    Dec 29, 2010 · This paper introduces and describes a solution to this famous problem using a new metaheuristic nature-inspired algorithm, called firefly algorithm.
  22. [22]
    A hybrid discrete firefly algorithm for multi-objective flexible job shop ...
    Jun 1, 2014 · In this paper, a hybrid discrete firefly algorithm is presented to solve the multi-objective flexible job shop scheduling problem with ...
  23. [23]
    (PDF) Firefly algorithm combined with Newton method to identify ...
    Mar 27, 2017 · Transient inverse heat conduction problems (IHCPs) are widely applied in aerospace, petrochemical,. and other industrial fields [1–3].
  24. [24]
  25. [25]
    (PDF) Fire Fly Based Feature Selection Approach - ResearchGate
    Aug 10, 2025 · The paper presents a new feature selection approach that combines the RST with nature inspired 'firefly' algorithm. ... UCI machine learning ...
  26. [26]
    (PDF) Training Feed-Forward Neural Networks Using Firefly Algorithm
    For training the back-propagation neural network, Nandy et al. introduced the nature-inspired firefly algorithm [32]. ...
  27. [27]
    A Firefly-Inspired Method for Protein Structure Prediction in Lattice ...
    The Firefly Algorithm (FA) was developed by X.S. Yang in 2008 [27,28], which was based on the flashing patterns and behaviour of tropical fireflies. FA is ...Missing: sources | Show results with:sources
  28. [28]
    Class center-based firefly algorithm for handling missing data
    Feb 23, 2021 · The adaptive search procedure performed in the Firefly algorithm is used to overcome the missing data in a dataset. In addition, the use of the ...Related Work · The Proposed Methods · Experiment Result
  29. [29]
    A hybrid SVR with the firefly algorithm enhanced by a logarithmic ...
    Sep 1, 2022 · 1) From the perspective of parameter optimization, we introduced the Lévy-flight firefly algorithm (LF-FA) and the logarithmic spiral firefly algorithm (LS-FA) ...
  30. [30]
  31. [31]
    Multiobjective firefly algorithm for continuous optimization
    Jan 24, 2012 · Multiobjective firefly algorithm for continuous optimization. Original Article; Published: 24 January 2012 ... Multi-objective Firefly Algorithm ...Missing: 2012 | Show results with:2012
  32. [32]
    A hybrid algorithm based on particle swarm optimization
    Controller Design using Hybrid Firefly and Particle Swarm optimization Algorithm ... February 2014 · Applied Soft Computing. Jinn-Tsong Tsai · Ching-I. Yang ...
  33. [33]
    A hybrid Firefly-Genetic Algorithm for the capacitated facility location ...
    Nov 1, 2014 · In this paper, a new hybrid optimization method called Hybrid Evolutionary Firefly-Genetic Algorithm is proposed, which is inspired by social behavior of ...
  34. [34]
  35. [35]
    Weighing Efficiency-Robustness in Supply Chain Disruption by Multi ...
    Multi-objective Firefly Algorithm. The heuristic algorithm is quite powerful in solving optimization problems, which has been well researched in numerous papers ...
  36. [36]
    [PDF] A Comparison Between the Firefly Algorithm and Particle Swarm ...
    Mar 1, 2013 · The Firefly Algorithm. The firefly algorithm (FA) is another swarm intelligence algorithm, developed by Xin-She. Yang [3]. This new ...<|control11|><|separator|>
  37. [37]
    [PDF] A comprehensive review of firefly algorithms - arXiv
    Dec 23, 2013 · The randomization parameter α was adapted. The algorithm main- tained only one swarm of fireflies. It was applied to a moving peaks bench ...
  38. [38]
    [PDF] A Systematic Review on Firefly Algorithm: Past, Present, and Future
    Apr 14, 2020 · This social life is dedicated to reproduction of fireflies and foraging. ... (7) become zero and the movement of firefly follows random walk (and ...
  39. [39]
    [PDF] The Firefly Algorithm and the Basic Ant Colony Algori
    This paper introduces two heuristic algorithms and illustrates the similarities and differences between the two algorithms; in the end, through the empirical ...
  40. [40]
    Firefly Algorithm: Recent Advances and Applications - arXiv
    Aug 18, 2013 · In this paper, we will briefly review the fundamentals of firefly algorithm together with a selection of recent publications.Missing: original | Show results with:original
  41. [41]
    (PDF) A Systematic Review on Firefly Algorithm: Past, Present, and ...
    ... advantages, disadvantages, and suggestions for future avenues of research on ... strengths and its limitations, such as computational demands and ...
  42. [42]
    (PDF) Scalable Parallel Clustering Approach for Large Data Using ...
    Apr 18, 2016 · In this paper a parallelization of firefly algorithm is introduced. Firefly algorithm is one of the latest bio-inspired algorithms and this is ...
  43. [43]
    Enhancing workflow efficiency with a modified Firefly Algorithm for ...
    Oct 21, 2024 · This hybrid model enables more efficient resource allocation and workload distribution, optimizing both performance and cost. However, the ...