将Numpy矩阵值转换为float32

Question

I want to enforce the values in the eigenvector matrix below to be of type float32 to save memory space instead of using float64 . However, it seems the way I did below eigenvectors.astype(np.float32) does not work. Is there a work around of that?

import networkx as nx
G = nx.Graph(edges)
L = nx.laplacian_matrix(G)
eigenvalues, eigenvectors.astype(np.float32) = np.linalg.eig(L.todense())

Answer 1

The np.linalg.eig function returns arrays of the same dtype as its parameter.

So you have two choices:

1. Try np.linalg.eig(L.todense().astype(np.float32)) . (This might overflow)
2. Convert the eigenvectors later.

The second option goes like this:

eigenvalues, eigenvectors = np.linalg.eig(L.todense())
eigenvectors = eigenvectors.astype(np.float32)