3

我有一个 3D 多边形图并且想要平滑 y 轴上的图（即我希望它看起来像“曲面图的切片”）。

考虑这个 MWE（取自这里）：

from mpl_toolkits.mplot3d import Axes3D
from matplotlib.collections import PolyCollection
import matplotlib.pyplot as plt
from matplotlib import colors as mcolors
import numpy as np
from scipy.stats import norm

fig = plt.figure()
ax = fig.gca(projection='3d')

xs = np.arange(-10, 10, 2)
verts = []
zs = [0.0, 1.0, 2.0, 3.0]

for z in zs:
    ys = np.random.rand(len(xs))
    ys[0], ys[-1] = 0, 0
    verts.append(list(zip(xs, ys)))

poly = PolyCollection(verts, facecolors=[mcolors.to_rgba('r', alpha=0.6),
                                         mcolors.to_rgba('g', alpha=0.6), 
                                         mcolors.to_rgba('b', alpha=0.6), 
                                         mcolors.to_rgba('y', alpha=0.6)])
poly.set_alpha(0.7)
ax.add_collection3d(poly, zs=zs, zdir='y')
ax.set_xlabel('X')
ax.set_xlim3d(-10, 10)
ax.set_ylabel('Y')
ax.set_ylim3d(-1, 4)
ax.set_zlabel('Z')
ax.set_zlim3d(0, 1)
plt.show()

现在，我想用正态分布替换四个图（理想情况下形成连续线）。

我在这里创建了分布：

def get_xs(lwr_bound = -4, upr_bound = 4, n = 80):
    """ generates the x space betwee lwr_bound and upr_bound so that it has n intermediary steps """
    xs = np.arange(lwr_bound, upr_bound, (upr_bound - lwr_bound) / n) # x space -- number of points on l/r dimension
    return(xs)

xs = get_xs()

dists = [1, 2, 3, 4]

def get_distribution_params(list_):
    """ generates the distribution parameters (mu and sigma) for len(list_) distributions"""
    mus = []
    sigmas = []
    for i in range(len(dists)):
        mus.append(round((i + 1) + 0.1 * np.random.randint(0,10), 3))
        sigmas.append(round((i + 1) * .01 * np.random.randint(0,10), 3))
    return mus, sigmas

mus, sigmas = get_distribution_params(dists)

def get_distributions(list_, xs, mus, sigmas):
    """ generates len(list_) normal distributions, with different mu and sigma values """
    distributions = [] # distributions

    for i in range(len(list_)):
        x_ = xs
        z_ = norm.pdf(xs, loc = mus[i], scale = sigmas[0])
        distributions.append(list(zip(x_, z_)))
        #print(x_[60], z_[60])

    return distributions

distributions = get_distributions(list_ = dists, xs = xs, mus = mus, sigmas = sigmas)

但是将它们添加到代码中（使用poly = PolyCollection(distributions, ...)并ax.add_collection3d(poly, zs=distributions, zdir='z')抛出ValueError( ValueError: input operand has more dimensions than allowed by the axis remapping) 我无法解决。

4

2 回答 2

2

该错误是由于传递distributions到zswherezs期望verts在PolyCollection具有形状MxNx2时传递给的对象zs具有形状M引起的。所以当它到达这个检查

cpdef ndarray broadcast_to(ndarray array, shape):
    # ...
    if array.ndim < len(shape):
        raise ValueError(
            'input operand has more dimensions than allowed by the axis '
            'remapping')
    # ...

在底层的 numpy 代码中，它失败了。我相信这是因为预期的维度数 ( )array.ndim小于( ) 的维度数。它期待一个 shape 数组，但接收一个 shape 数组。zslen(shape)(4,)(4, 80, 2)

这个错误可以通过使用正确形状的数组来解决 - 例如zs来自原始示例或dists来自您的代码。使用zs=dists和调整轴范围到[0,5]for x, y, 并z给出

在此处输入图像描述

这看起来有点奇怪，原因有两个：

  1. 有一个错字，z_ = norm.pdf(xs, loc = mus[i], scale = sigmas[0])其中给所有分布相同的西格玛，它应该是z_ = norm.pdf(xs, loc = mus[i], scale = sigmas[i])
  2. 观察几何：分布以正xz平面为基础，这也是我们正在观察的平面。

通过更改查看几何图形ax.view_init将产生更清晰的图：

在此处输入图像描述

编辑

这是生成所示图的完整代码，

from mpl_toolkits.mplot3d import Axes3D
from matplotlib.collections import PolyCollection
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors
import numpy as np
from scipy.stats import norm

np.random.seed(8)
def get_xs(lwr_bound = -4, upr_bound = 4, n = 80):
    return np.arange(lwr_bound, upr_bound, (upr_bound - lwr_bound) / n)

def get_distribution_params(list_):
    mus = [round((i+1) + 0.1 * np.random.randint(0,10), 3) for i in range(len(dists))]
    sigmas = [round((i+1) * .01 * np.random.randint(0,10), 3) for i in range(len(dists))]
    return mus, sigmas

def get_distributions(list_, xs, mus, sigmas):
    return [list(zip(xs, norm.pdf(xs, loc=mus[i], scale=sigmas[i] if sigmas[i] != 0.0 
            else 0.1))) for i in range(len(list_))]

dists = [1, 2, 3, 4]
xs = get_xs()
mus, sigmas = get_distribution_params(dists)
distributions = get_distributions(dists, xs, mus, sigmas)

fc = [mcolors.to_rgba('r', alpha=0.6), mcolors.to_rgba('g', alpha=0.6), 
      mcolors.to_rgba('b', alpha=0.6), mcolors.to_rgba('y', alpha=0.6)]

poly = PolyCollection(distributions, fc=fc)
fig = plt.figure()
ax = fig.gca(projection='3d')
ax.add_collection3d(poly, zs=np.array(dists).astype(float), zdir='z')
ax.view_init(azim=115)
ax.set_zlim([0, 5])
ax.set_ylim([0, 5])
ax.set_xlim([0, 5])

我基于您在问题中提供的代码，但为了简洁起见并与通常的样式更加一致，做了一些修改。

注意- 您给出的示例代码np.random.seed()将   根据.norm.pdfscale = sigma[i] if sigma[i] != 0.0 else 0.1

于 2020-01-15T23:52:57.833 回答
1

使用ax.add_collection3d(poly, zs=dists, zdir='z')而不是ax.add_collection3d(poly, zs=distributions, zdir='z')应该解决问题。

此外，您可能想要更换

def get_xs(lwr_bound = -4, upr_bound = 4, n = 80):
    """ generates the x space betwee lwr_bound and upr_bound so that it has n intermediary steps """
    xs = np.arange(lwr_bound, upr_bound, (upr_bound - lwr_bound) / n) # x space -- number of points on l/r dimension
    return(xs)

xs = get_xs()

经过

xs = np.linspace(-4, 4, 80)

另外，我相信scale = sigmas[0]实际上应该scale = sigmas[i]在行中

z_ = norm.pdf(xs, loc = mus[i], scale = sigmas[0])

最后，我相信您应该适当地调整和xlim，因为您在与参考代码进行比较时交换了绘图的和尺寸并更改了它的比例。ylimzlimyz

于 2020-01-15T03:21:17.150 回答