迭代地减去数组中的值 [英] Iteratively subtract values in array

问题描述

从简单的numpy数组开始，例如:

Starting from a simple numpy array like:

a = np.array([1,1,0,2,1,0])

我的目标是从中迭代地减去值直到某个阈值.例如，考虑包含用户数量的a.在这种情况下，向5个用户分配:

my goal is to iteratively subtract values from this till a certain threshold. For instance, think of a including number of users. In this case 5 users are distributed with:

d = a/a.sum()

现在，我想从这个初始分布中减去4个用户，并在结果数组中始终保持值> 0.我可以构造一个随机数组减去:

Now I want to subtract 4 users at the time from this initial distribution, keeping always values > 0 in the resulting array. I can construct a random array to subtract with:

b = np.random.multinomial(4,d)

生成(一次运行)的

:

that generates (in a single run):

array([0, 1, 0, 3, 0, 0])

类似于a-b的结果导致:

array([ 1,  0,  0, -1,  1,  0])

如何约束生成的数组在结果(a-b)操作中永远不会产生负值?到目前为止，我认为是从另一端开始做事，生成了以下内容的随机分布:

How can I constraint that the generated array never produces a negative value in the resulting (a-b) operation? So far I thought in doing the things from the other side, generating a random distribution of:

r = np.random.multinomial(total users - deleted users,d)

根据用户d的初始分布，但是由于我必须在向量上应用一些指标，因此后一种方法的结果可能会有所不同.

according to the initial distribution of users d, but since I have to apply some metrics on the vector, results might vary in the latter approach.

推荐答案

如果您想将a.sum()随机递减n而不使a的任何元素变为负数，这是做到这一点的一种方法:

If you want to randomly decrement a.sum() by n without any element of a becoming negative, this is one way to do it:

def rnd_decrement(a, n):
    c = np.cumsum(np.r_[0, a])
    if n < c[-1]:
        r = np.random.choice(np.arange(c[-1]) + 1, n, replace = False)
        d = np.sum(r[:,None] <= c[None,:], axis=0)
        return np.diff(c-d)
    else:
        return np.zeros_like(a)

这篇关于迭代地减去数组中的值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！