(令人惊讶的挑战?)Numpy矢量化

时间:2020-04-17 08:44:29

标签: python numpy vectorization

我想在我的代码中找到一种方法避免循环。我需要实现以下公式,起初很简单:

formula

换句话说:索引的列表被解析为 I 。对于 I 中指定的每个索引,需要减去数组 x 中所有后续索引的值。对减去的值进行一些计算。总结一切。完成。

我当前的代码:

def loss(x, indices):
    """ 
    Args:
        x: array_like, dtype=float
        indices: array_like, dtype=int

    Example:
        >>> x = np.array([0.3, 0.5, 0.2, 0.1, 1.2, 2.4, 2.8, 1.5, 3.2])
        >>> indices = np.array([0, 2, 3, 6])
        >>> print(loss(x, indices))
        21.81621815885847
    """

    total = 0.0
    for index in indices:
        # Broadcasting here, as all values from all following indices have
        # to be subtracted from the value at the given i index.
        difference = x[index] - x[index + 1:]

        # Sum all up
        log_addition = 1.0 + np.log(np.abs(difference))
        total += np.sum(log_addition)

    return total

具有挑战性的部分是'i'索引在输出范围内随机分布。有什么想法吗?

1 个答案:

答案 0 :(得分:4)

这里有一个基于NumPy的矢量化-

mask = indices[:,None] < np.arange(len(x))
v = x[indices,None] - x
vmasked = v[mask]
log_addition = np.log(np.abs(vmasked))
out = log_addition.sum() + mask.sum()

或者,根据对数定律,我们可以将最后两个步骤替换为-

out = np.log(np.prod(np.abs(vmasked))).sum() + mask.sum()

abs推出,以使其在标量上运行,它将是-

out = np.log(np.abs(np.prod(vmasked))).sum() + mask.sum()

同样,我们可以将multi-coresnumexpr一起使用-

import numexpr as ne
out = np.log(np.abs(ne.evaluate('prod(vmasked)'))) + mask.sum()

如果您发现甚至v中也有太多不需要的元素,我们可以直接使用-

转到vmasked
xi = x[indices]
x2D = np.broadcast_to(x, (len(indices),len(x)))
vmasked = np.repeat(xi,mask.sum(1))-x2D[mask]
相关问题