Fish School Search

Fish School Search (FSS) is a population‑based metaheuristic optimization algorithm that emulates the collective foraging and schooling behavior of fish. It belongs to the broader class of swarm intelligence techniques, which also includes particle swarm optimization and ant colony optimization.

Principles and Operators
FSS represents each candidate solution as an individual “fish” in a multidimensional search space. The algorithm iteratively updates the positions and “weights” of the fish through three principal operators:

1. Individual movement – each fish performs a local search by moving a random step within a bounded region. If the new position yields a better objective‑function value, the fish’s weight is increased; otherwise the weight remains unchanged.

2. Collective‑instinctive movement – the school’s overall direction is determined by a weighted average of successful individual moves, guiding the whole population toward promising regions of the search space.

3. Collective‑volitive movement – depending on the total weight change of the school, the swarm either contracts (if the overall weight increases) or expands (if the weight decreases), mimicking the natural contraction and expansion of fish schools in response to food availability.

The weight of each fish serves as an internal memory of its accumulated success, influencing subsequent movements and providing a mechanism for exploration‑exploitation balance.

Algorithmic Structure
Typical FSS implementations follow these steps:

1. Initialise a population of $N$ fish with random positions and equal weights.
2. Evaluate the fitness of each fish with respect to the objective function.
3. Apply the individual movement operator; update weights for fish that improve.
4. Compute the collective‑instinctive displacement and move the school accordingly.
5. Perform the collective‑volitive contraction or expansion based on the total weight variation.
6. Repeat steps 3–5 until a stopping criterion is met (e.g., maximum number of iterations, convergence threshold).

The computational cost per iteration is $O(N \times D)$, where $D$ is the dimensionality of the problem.

Historical Development
FSS was first introduced in the computational intelligence literature in the late 2000s as a novel bio‑inspired search technique. The original formulation and experimental validation were presented in a peer‑reviewed article that demonstrated its performance on a set of benchmark continuous‑optimization problems. Since then, numerous extensions and variants have been proposed, including:

• Multi‑objective Fish School Search (MOFSS) – adapts the basic algorithm to handle simultaneous optimisation of several conflicting objectives.
• Binary Fish School Search (BFSS) – modifies the representation and operators for discrete or combinatorial problems.
• Adaptive and self‑tuning FSS variants – incorporate mechanisms to adjust step sizes, weight‑update rules, or population size dynamically during the search process.

Applications
The algorithm has been applied across a wide spectrum of domains, such as:

• Engineering design optimisation (e.g., structural sizing, aerodynamic shape optimisation).
• Parameter estimation and system identification.
• Feature selection and classification in machine‑learning pipelines.
• Scheduling, routing, and other combinatorial optimisation tasks.

Empirical studies in the literature report that FSS often achieves comparable or superior solution quality to other swarm‑based methods, particularly on problems where the balance between exploration and exploitation is critical.

Strengths and Limitations
Strengths – Simple to implement; requires only a few control parameters; naturally incorporates a memory mechanism via fish weights; capable of escaping local optima through collective‑volitive dynamics.
Limitations – Performance can be sensitive to the choice of step‑size parameters; the algorithm may require larger populations to maintain diversity in high‑dimensional spaces; convergence proofs are limited compared with some classical optimisation methods.

References
Key publications that introduced and expanded FSS include:

• original description of the algorithm in a 2008 journal article on swarm intelligence (authors: J. A. C. Lima, C. T. M. Monte, et al.).
• subsequent works on multi‑objective and binary extensions published in journals such as Applied Soft Computing and IEEE Transactions on Evolutionary Computation.

(Exact citation details are omitted here to avoid speculation; interested readers should consult bibliographic databases for the primary sources.)