使用matplotlib

Question

我的数据看起来像（例子）

x y  d
0 0 -2
1 0  0
0 1  1 
1 1  3

我想把它变成一个coloumap图，看起来像其中之一：

其中x和y在表中，颜色由'd'给出。但是，我想为每个数字预定颜色，例如：

-2 - orange
 0 - blue
 1 - red
 3 - yellow

不一定是这些颜色，但我需要将一个数字用于颜色，数字不是按顺序或顺序排列，只是一组五个或六个随机数，它们在整个数组中重复。

任何想法，我都没有代码，因为我不知道从哪里开始。不过我看过这里的例子如下：

Matplotlib python change single color in colormap

但是，它们仅显示如何定义颜色，而不显示如何将这些颜色与特定值相关联。

Answer 1

事实证明这比我想象的要难，所以也许某人有更简单的方法来做这件事。

由于我们需要创建数据图像，因此我们将它们存储在2D数组中。然后，我们可以将数据映射到整数0 .. number of different data values，并为每个整数分配颜色。原因是我们希望最终的色图是等间距的。所以

值-2 - ＆gt;整数0 - ＆gt;颜色orange
值0 - ＆gt;整数1 - ＆gt;颜色blue 等等。

具有间隔良好的整数，我们可以在新创建的整数值的图像上使用ListedColormap。

import matplotlib.pyplot as plt
import numpy as np
import matplotlib.colors

# define the image as a 2D array
d = np.array([[-2,0],[1,3]])

# create a sorted list of all unique values from d
ticks = np.unique(d.flatten()).tolist()
# create a new array of same shape as d
# we will later use this to store values from 0 to number of unique values
dc = np.zeros(d.shape)
#fill the array dc
for i in range(d.shape[0]):
    for j in range(d.shape[1]):
        dc[i,j] = ticks.index(d[i,j])

# now we need n (= number of unique values) different colors        
colors= ["orange", "blue", "red", "yellow"]
# and put them to a listed colormap
colormap =  matplotlib.colors.ListedColormap(colors)


plt.figure(figsize=(5,3))
#plot the newly created array, shift the colorlimits, 
# such that later the ticks are in the middle
im = plt.imshow(dc, cmap=colormap, interpolation="none", vmin=-0.5, vmax=len(colors)-0.5)
# create a colorbar with n different ticks
cbar = plt.colorbar(im, ticks=range(len(colors)) )
#set the ticklabels to the unique values from d
cbar.ax.set_yticklabels(ticks)
#set nice tickmarks on image
plt.gca().set_xticks(range(d.shape[1]))
plt.gca().set_yticks(range(d.shape[0]))

plt.show()

由于可能不直观清楚如何将数组d置于使用imshow绘制所需的形状中，即作为2D数组，这里有两种转换输入数据列的方法：

import numpy as np

x = np.array([0,1,0,1])
y  = np.array([ 0,0,1,1])
d_original = np.array([-2,0,1,3])

#### Method 1 ####
# Intuitive method. 
# Assumption: 
#    * Indexing in x and y start at 0 
#    * every index pair occurs exactly once.
# Create an empty array of shape (n+1,m+1) 
#   where n is the maximum index in y and
#         m is the maximum index in x
d = np.zeros((y.max()+1 , x.max()+1), dtype=np.int) 
for k in range(len(d_original)) :
    d[y[k],x[k]] = d_original[k]  
print d

#### Method 2 ####
# Fast method
# Additional assumption: 
#  indizes in x and y are ordered exactly such
#  that y is sorted ascendingly first, 
#  and for each index in y, x is sorted. 
# In this case the original d array can bes simply reshaped
d2 = d_original.reshape((y.max()+1 , x.max()+1))
print d2