有效地填充具有许多if语句的多维数组 [英] Fill a multidimensional array efficiently that have many if else statements

问题描述

我想以一种特定且有效的方式填充4dim numpy数组.因为我不太了解，所以我开始用if else语句编写代码，但这看起来并不好，可能很慢，而且我也不确定我是否考虑过每种组合.这是我停止写下的代码:

I want to fill an 4dim numpy array in a specific and efficient way. Because I don't know better I startet to write the code with if else statements, but that doesn't look nice, is probably slow and I also can not be really sure if I thought about every combination. Here is the code which I stopped writing down:

sercnew2 = numpy.zeros((gn, gn, gn, gn))
for x1 in range(gn):
    for x2 in range(gn):
        for x3 in range(gn):
            for x4 in range(gn):
                if x1 == x2 == x3 == x4: 
                    sercnew2[x1, x2, x3, x4] = ewp[x1]
                elif x1 == x2 == x3 != x4:
                    sercnew2[x1, x2, x3, x4] = ewp[x1] * ewp[x4]
                elif x1 == x2 == x4 != x3:
                    sercnew2[x1, x2, x3, x4] = ewp[x1] * ewp[x3]
                elif x1 == x3 == x4 != x2:
                    sercnew2[x1, x2, x3, x4] = ewp[x1] * ewp[x2]
                elif x2 == x3 == x4 != x1:
                    sercnew2[x1, x2, x3, x4] = ewp[x2] * ewp[x1]
                elif x1 == x2 != x3 == x4:
                    sercnew2[x1, x2, x3, x4] = ewp[x1] * ewp[x3]
                elif ... many more combinations which have to be considered

所以基本上应该发生的是，如果所有变量(x1，x2，x3，x4)互不相同，则条目将为:

So basically what should happen is, that if all variables (x1, x2, x3, x4) are different from each other, the entry would be:

sercnew2[x1, x2, x3, x4] = ewp[x1]* ewp[x2] * ewp[x3] * ewp[x4]

现在，如果说变量x2和x4相同，则:

Now if lets say the variable x2 and x4 is the same then:

sercnew2[x1, x2, x3, x4] = ewp[x1]* ewp[x2] * ewp[x3]

其他示例可以在上面的代码中看到.基本上，如果两个或多个变量相同，那么我只考虑其中的一个.我希望模式是明确的.否则，请让我注意一下，我将尝试更好地表达我的问题.我很确定，有一种更加智能的方法可以做到这一点.希望您知道得更多，并提前致谢:)

Others examples can be seen in the code above. Basically if two or more variables are the same, then I only consider on of them. I hope the pattern is clear. Otherwise please let me note and I will try to express my problem better. I am pretty sure, that there is a much more intelligent way to do it. Hope you know better and thanks in advance :)

推荐答案

真的希望我已经知道了！这是向量化方法-

Really hoping that I have got it! Here's a vectorized approach -

from itertools import product

n_dims = 4 # Number of dims

# Create 2D array of all possible combinations of X's as rows
idx = np.sort(np.array(list(product(np.arange(gn), repeat=n_dims))),axis=1)

# Get all X's indexed values from ewp array
vals = ewp[idx]

# Set the duplicates along each row as 1s. With the np.prod coming up next,
#these 1s would not affect the result, which is the expected pattern here.
vals[:,1:][idx[:,1:] == idx[:,:-1]] = 1

# Perform product along each row and reshape into multi-dim array
out = vals.prod(1).reshape([gn]*n_dims)

这篇关于有效地填充具有许多if语句的多维数组的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！