import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D


INF = 1e5

def plot_cost_func():
  """畫(huà)出適應(yīng)度函數(shù)"""
  fig = plt.figure()
  ax = Axes3D(fig)
  X = np.arange(-4, 4, 0.25)
  Y = np.arange(-4, 4, 0.25)
  X, Y = np.meshgrid(X, Y)
  Z = (X**2 + Y**2) - 10
  ax.plot_surface(X, Y, Z, rstride=1, cstride=1, cmap='rainbow')
  plt.show()

def fitness(x):
  return x[0]**2 + x[1]**2 - 10

class PSOSolver(object):
  def __init__(self, n_iter, weight=0.5, c1=2, c2=2, n_particle=5):
    self.n_iter = n_iter
    self.weight = weight
    self.c1 = c1
    self.c2 = c2
    self.n_particle = n_particle
    self.gbest = np.random.rand(2)
    # gbest 對(duì)應(yīng)的函數(shù)值
    self.gbest_fit = fitness(self.gbest)
    # 將位置初始化到 [-5, 5]
    self.location = 10 * np.random.rand(n_particle, 2) - 5
    # 將速度初始化到 [-1, 1]
    self.velocity = 2 * np.random.rand(n_particle, 2) - 1
    self.pbest_fit = np.tile(INF, n_particle)
    self.pbest = np.zeros((n_particle, 2))
    # 記錄每一步的最優(yōu)值
    self.best_fitness = []
  
  def new_velocity(self, i):
    r = np.random.rand(2, 2)
    v = self.velocity[i]
    x = self.location[i]
    pbest = self.pbest[i]
    return self.weight * v + self.c1 * r[0] * (pbest - x) + \
        self.c2 * r[1] * (self.gbest - x)

  def solve(self):
    for it in range(self.n_iter):
      for i in range(self.n_particle):
        v = self.new_velocity(i)
        x = self.location[i] + v
        fit_i = fitness(x)
        if fit_i < self.pbest_fit[i]:
          self.pbest_fit[i] = fit_i
          self.pbest[i] = x
          if fit_i < self.gbest_fit:
            self.gbest_fit = fit_i
            self.gbest = x
        self.velocity[i] = v
        self.location[i] = x
      self.best_fitness.append(self.gbest_fit)

  
if __name__ == '__main__':
  plot_cost_func()
  n_iter = 20
  s = PSOSolver(n_iter)
  s.solve()
  print(s.gbest_fit)
  plt.title("Fitness Curve")
  plt.xlabel("iter")
  plt.ylabel("fitness")
  plt.plot(np.arange(n_iter), np.array(s.best_fitness))
  plt.show()