在 PyTorch 中用向量替换对角线元素

问题描述

我一直在到处寻找与 PyTorch 类似的东西，但我找不到任何东西.

I have been searching everywhere for something equivalent of the following to PyTorch, but I cannot find anything.

L_1 = np.tril(np.random.normal(scale=1., size=(D, D)), k=0)
L_1[np.diag_indices_from(L_1)] = np.exp(np.diagonal(L_1))

我想没有办法使用 Pytorch 以如此优雅的方式替换对角线元素.

I guess there is no way to replace the diagonal elements in such an elegant way using Pytorch.

推荐答案

我认为目前还没有实现这样的功能.但是，您可以使用 mask 实现相同的功能，如下所示.

I do not think that such a functionality is implemented as of now. But, you can implement the same functionality using mask as follows.

# Assuming v to be the vector and a be the tensor whose diagonal is to be replaced
mask = torch.diag(torch.ones_like(v))
out = mask*torch.diag(v) + (1. - mask)*a

因此，您的实现将类似于

So, your implementation will be something like

L_1 = torch.tril(torch.randn((D, D)))
v = torch.exp(torch.diag(L_1))
mask = torch.diag(torch.ones_like(v))
L_1 = mask*torch.diag(v) + (1. - mask)*L_1

没有 numpy 优雅，但也不算太糟糕.

Not as elegant as numpy, but not too bad either.

这篇关于在 PyTorch 中用向量替换对角线元素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编程学习网！