Particle Swarm Optimization in Python

pso.py

import numpy as np

def basic_pso(cost_function, bounds, num_particles, num_iterations):
    # Initialize the positions and velocities of the particles
    positions = np.random.uniform(bounds[0], bounds[1], (num_particles, len(bounds[0])))
    velocities = np.zeros((num_particles, len(bounds[0])))
    # Initialize the best position and cost for each particle
    best_positions = positions.copy()
    best_costs = np.array([cost_function(p) for p in positions])
    # Initialize the global best position and cost
    global_best_position = best_positions[np.argmin(best_costs)]
    global_best_cost = np.min(best_costs)
    # Set the constants for the velocity update formula
    c1 = 2.05
    c2 = 2.05
    # Run the iterations of the algorithm
    for _ in range(num_iterations):
        for i in range(num_particles):
            # Update the velocity and position of the particle
            velocities[i] = (0.7298 * velocities[i]) + (c1 * np.random.rand() * (best_positions[i] - positions[i])) + (c2 * np.random.rand() * (global_best_position - positions[i]))
            positions[i] += velocities[i]
            # Update the best position and cost for the particle if a better solution is found
            cost = cost_function(positions[i])
            if cost < best_costs[i]:
                best_positions[i] = positions[i]
                best_costs[i] = cost
            # Update the global best position and cost if a better solution is found
            if cost < global_best_cost:
                global_best_position = positions[i]
                global_best_cost = cost
    # Return the final global best position and cost
    return global_best_position, global_best_cost

def fitness(x):
    return x**2

result = basic_pso(fitness, ([-10.0], [10.0]), 25, 100)

print("Global best position: " + str(result[0]))
print("Global best fitness:  " + str(result[1]))