vfo-optimizer 0.0.2

Van Fish Optimization (VFO) algorithm

Van Fish Optimization (VFO) Algorithm

Van Fish Optimization (VFO) is a meta-heuristic optimization algorithm inspired by the unique migration behavior of the Pearl Mullet (Alburnus tarichi) living in Lake Van, Turkey. These fish swim against the current and jump over obstacles during migration, behaviours which VFO models to solve complex optimization problems.

Key features include:

• Directional Search: Mimics swimming against the current towards a target.
• Local Escape: Uses a jumping mechanism (Lévy flight) to escape local minima, modeled after fish jumping obstacles.
• Adaptive Parameters: Uses energy-based and time-varying parameters to balance exploration and exploitation.

---

Installation

You can install the package via pip:

pip install vfo-optimizer

---

Usage

Here is a simple example of how to use VFO to minimize the Sphere function:

import numpy as np
import VFO

# 1. Define your objective function (must take a 1D array and return a scalar)
def sphere_function(x):
    return np.sum(x**2)

# 2. Define problem bounds (lower and upper limits for each dimension)
dim = 10
bounds = [[-100, 100]] * dim

# 3. Initialize the VFO optimizer
optimizer = VFO.VFO(
    objective_func=sphere_function,
    bounds=np.array(bounds),
    num_agents=30,      # Population size
    max_iter=100,       # Maximum iterations
    dimension=dim,      # Problem dimension
    verbose=True        # Print progress
)

# 4. Run optimization
best_position, best_fitness, history = optimizer.optimize()

print(f"Best Fitness Value: {best_fitness}")
# print(f"Best Position: {best_position}")

---

Hyperparameters

When initializing the VFO class, you can tune the following parameters:

Parameter	Type	Default	Description
objective_func	callable	Required	The objective function you want to minimize.
bounds	array	Required	A list of [min, max] tuples for each dimension.
dimension	int	2	The number of dimensions (variables) in the problem.
num_agents	int	30	Population Size (N). Higher values check more areas but increase computation time.
max_iter	int	100	Maximum Iterations. How many times the population updates itself.
stagnation_threshold	int	5	Jump Trigger. If an agent doesn't improve for this many steps, it performs a Lévy flight to jump out of local minima.
verbose	bool	False	If True, prints the best fitness value at each iteration.

---

Theoretical Model

The algorithm mathematically models the fish behavior using three main components:

1. Position Update (Swimming): Fish move towards the leader ($X_{best}$) but are affected by the current ($A$) and randomness ($B$). $$X_{new} = X_{current} + A \cdot (X_{best} - X_{current}) + B \cdot R$$

2. Adaptive Coefficients:

   • $A$ (Current Resistance): Decreases linearly from 2 to 0. Promotes exploration early on and exploitation later.
   • $B$ (Randomness): Decreases over time to reduce random scattering as the target is approached.

3. Lévy Flight (Jumping): If a fish (agent) cannot improve its position for a certain number of steps (stagnation_threshold), it assumes it is stuck at an obstacle (local minimum) and performs a long-distance jump using a Lévy distribution.

---