我想编写一个简单的程序来提高我对这类编程的知识。
我找到了一个非常有用的库pyeasyGA，并尝试使用graphics.py创建一个简单的程序，该程序从随机生成的“passs”序列创建一个收敛到一个点的序列。
这就是它的工作方式:
def create_individual(data):
a = [(randint(0,5),randint(0,5)) for n in range(len(data))]
print(a)
return a
该函数创建了一系列的传递，因为graphics.py库允许您通过给对象提供要移动的像素数来移动对象。那是我的“个人”。
为了计算适合度，我使用了以下方法:
def fitness(individual, data):
totX=0
totY=0
for elem in individual:
totX+=elem[0]
totY+=elem[1]
tot = (totX,totY)
return distEuclidea(arrivo, tot)
def distEuclidea(p1,p2):
x1 = p1[0]
y1 = p1[1]
x2 = p2[0]
y2 = p2[1]
return ((x2-x1)**2+(y2-y1)**2)**(1/2)
此功能计算距所需到达点的距离。
经过这些测试后，该程序会生成很多代，并采用适应性最低的个人，但它不起作用。
它不会发展。通过的每个序列似乎都是随机生成的。
有谁可以帮助我吗？
Here's the full code
编辑:
该程序似乎有效。唯一的问题是几代人。
最佳答案
我发现您的健身功能最难理解。而不是平均拐角或找到中心，它将拐角加起来然后找到距离。什么是几何解释？
此外，您的代码还引用了ga.logGenerations，它不属于当前pyeasyga 0.3.1版本。
以下是我想您所要求的近似值。如果不可行，请通过示例和/或图表来增加您的解释:
from time import sleep
from random import randint
from itertools import cycle
from graphics import *
from pyeasyga import pyeasyga
NUMBER_OF_RECTANGLES = 4 # make one more than what you want to see
NUMBER_OF_POINTS = 2
arrivo = (90, 90)
colori = ["red", "green", "blue", "cyan", "magenta", "yellow"]
X, Y = 0, 1
def distEuclidea(p1, p2):
x1, y1 = p1
x2, y2 = p2
return ((x2 - x1) ** 2 + (y2 - y1) ** 2) ** 0.5
def create_individual(colors):
color = next(colors)
while color in rectangles and rectangles[color] is None: # skip over deleted rectangle
color = next(colors)
if color in rectangles:
rectangle = rectangles[color]
p1, p2 = rectangle.getP1(), rectangle.getP2()
points = [[p1.getX(), p1.getY()], [p2.getX(), p2.getY()]]
else:
points = [[randint(0, 20), randint(0, 20)] for _ in range(NUMBER_OF_POINTS)]
rectangle = Rectangle(*[Point(x, y) for x, y in points])
rectangle.setOutline(color)
rectangle.draw(win)
rectangles[color] = rectangle
return [color, points]
def fitness(individual, colors):
_, points = individual
rectangle = Rectangle(*[Point(x, y) for x, y in points])
center = rectangle.getCenter()
return distEuclidea(arrivo, (center.getX(), center.getY()))
def mutate(individual):
_, points = individual
mutate_index = randint(0, NUMBER_OF_POINTS - 1)
points[mutate_index][X] += randint(-1, 1)
points[mutate_index][Y] += randint(-1, 1)
def is_point_inside_rectangle(point, rectangle):
p1, p2 = rectangle.getP1(), rectangle.getP2()
return min(p1.getX(), p2.getX()) < point.getX() < max(p1.getX(), p2.getX()) and \
min(p1.getY(), p2.getY()) < point.getY() < max(p1.getY(), p2.getY())
win = GraphWin("Genetic Graphics", 500, 500)
win.setCoords(0, 0, 100, 100)
rectangles = {}
color_generator = cycle(colori[0:NUMBER_OF_RECTANGLES])
arrivoC = Circle(Point(*arrivo), 1)
arrivoC.setFill("orange")
arrivoC.draw(win)
number_of_rectangles = NUMBER_OF_RECTANGLES
while True:
ga = pyeasyga.GeneticAlgorithm(color_generator, \
elitism=False, \
maximise_fitness=False, \
crossover_probability=0.0, \
population_size=number_of_rectangles)
ga.create_individual = create_individual
ga.fitness_function = fitness
ga.mutate_function = mutate
ga.run()
for member in ga.last_generation():
my_fitness, (my_color, my_points) = member
if rectangles[my_color] is None:
continue # skip over deleted rectangle
rectangle = Rectangle(*[Point(x, y) for x, y in my_points])
rectangle.setOutline(my_color)
rectangle.draw(win)
rectangles[my_color] = rectangle
if is_point_inside_rectangle(arrivoC.getCenter(), rectangle):
rectangles[my_color] = None # delete finished rectangle
number_of_rectangles -= 1
if number_of_rectangles < 2:
break
sleep(0.1)
for value in rectangles.values():
if value is not None:
value.undraw() # delete unfinished rectangle
win.getMouse()
win.close()
上面是粗糙的代码(例如，它并不总是保持通用的域点和矩形独立于graphics.py点和矩形。)但是它应该给您一些试验的方法:
它在窗口的左下角创建矩形，遗传算法会在右上角向目标突变，在矩形到达目标时将其删除。
我的代码的部分复杂性在于pyeasyga并没有提供一个功能钩子(Hook)来可视化每一代发生的事情。更好的方法可能是将pyeasyga子类化以添加这样的钩子(Hook)，以简化代码的逻辑。