psoco 0.0.8

partical swarm optimization constraint optimization solver

Particle Swarm Optimization Constraint Optimization Solver

Arguments

Name	Type	Default Value
particle_size	int	2000
max_iter	int	1000
sol_size	int	7
fitness	function	null
constraints	a list of functions	null

Usage

Transform constraints, it becomes:

Note: In order to faster search optimal solutions, please initialize solutions with specific low and high.

import psoco
import math 

def objective(x):
    '''create objectives based on inputs x as 2D array'''
    return (x[:, 0] - 2) ** 2 + (x[:, 1] - 1) ** 2 


def constraints1(x):
    '''create constraint1 based on inputs x as 2D array'''
    return x[:, 0] - 2 * x[:, 1] + 1 


def constraints2(x):
    '''create constraint2 based on inputs x as 2D array'''
    return - (x[:, 0] - 2 * x[:, 1] + 1)


def constraints3(x):
    '''create constraint3 based on inputs x as 2D array'''
    return x[:, 0] ** 2 / 4. + x[:, 1] ** 2 - 1

def new_penalty_func(k):
    '''Easy Problem can use \sqrt{k}'''
    return math.sqrt(k)
    
constraints = [constraints1, constraints2, constraints3]
num_runs = 10
# random parameters lead to variations, so run several time to get mean
for _ in range(num_runs):
    pso = psoco.PSOCO(sol_size=2, fitness=objective, constraints=constraints)
    pso.h = new_penalty_func
    pso.init_Population(low=0, high=1) # x并集的上下限，默认为0和1
    pso.solve()
    # best solutions
    x = pso.gbest.reshape((1, -1))

Reference

• Particle Swarm Optimization Method for Constrained Optimization Problems