如何在NumPy中将三角矩阵转换为正方形? [英] How to convert triangle matrix to square in NumPy?
问题描述
我正在对一个冗余的完整矩阵进行一些计算(即可以是一个三角形矩阵而不会丢失任何信息).我意识到我只能计算三角形的下部以获得更快的结果.完成后如何将下三角投影到上三角?
I'm doing some computations on a full matrix that is redundant (i.e. can be a triangle matrix without losing info). I realized I can compute only the lower portion of the triangle for faster results. How can I project the lower triangle into the upper once I'm done?
换句话说,我该如何反转np.tril
方法?
In other words, how can I reverse the np.tril
method?
print DF_var.as_matrix()
# [[1 1 0 1 1 1 0 1 0 0 0]
# [1 1 1 1 1 0 1 0 1 1 1]
# [0 1 1 0 0 0 0 0 0 0 0]
# [1 1 0 1 0 0 0 0 0 0 0]
# [1 1 0 0 1 0 0 0 0 0 0]
# [1 0 0 0 0 1 1 0 0 0 0]
# [0 1 0 0 0 1 1 0 0 0 0]
# [1 0 0 0 0 0 0 1 1 0 0]
# [0 1 0 0 0 0 0 1 1 0 0]
# [0 1 0 0 0 0 0 0 0 1 0]
# [0 1 0 0 0 0 0 0 0 0 1]]
print np.tril(DF_var.as_matrix())
# [[1 0 0 0 0 0 0 0 0 0 0]
# [1 1 0 0 0 0 0 0 0 0 0]
# [0 1 1 0 0 0 0 0 0 0 0]
# [1 1 0 1 0 0 0 0 0 0 0]
# [1 1 0 0 1 0 0 0 0 0 0]
# [1 0 0 0 0 1 0 0 0 0 0]
# [0 1 0 0 0 1 1 0 0 0 0]
# [1 0 0 0 0 0 0 1 0 0 0]
# [0 1 0 0 0 0 0 1 1 0 0]
# [0 1 0 0 0 0 0 0 0 1 0]
# [0 1 0 0 0 0 0 0 0 0 1]]
如何将其转换回完整矩阵?
How to convert it back to a full matrix?
推荐答案
假定A
作为输入数组,下面列出了几种方法.
Assuming A
as the input array, few methods are listed below.
方法#1::在A
-
np.triu(A.T,1) + A
方法2:避免np.triu
进行A.T和A之间的直接求和,然后索引以设置对角线元素-
Approach #2 : Avoid np.triu
with direct summation between A.T and A and then indexing to set diagonal elements -
out = A.T + A
idx = np.arange(A.shape[0])
out[idx,idx] = A[idx,idx]
方法3::与上一个方法相同,但使用内置索引进行压缩-
Approach #3 : Same as previous one, but compact using in-builts for indexing -
out = A.T + A
np.fill_diagonal(out,np.diag(A))
方法4:与上一个方法相同,但具有布尔索引来设置对角线元素-
Approach #4 : Same as previous one, but with boolean indexing to set diagonal elements -
out = A.T + A
mask = np.eye(out.shape[0],dtype=bool)
out[mask] = A[mask]
方法5:对np.where
-
np.where(np.eye(A.shape[0],dtype=bool),A,A.T+A)
方法6::对所有具有np.where
-
np.where(np.triu(np.ones(A.shape[0],dtype=bool),1),A.T,A)
运行时测试
功能-
def func1(A):
return np.triu(A.T,1) + A
def func2(A):
out = A.T + A
idx = np.arange(A.shape[0])
out[idx,idx] = A[idx,idx]
return out
def func3(A):
out = A.T + A
np.fill_diagonal(out,np.diag(A))
return out
def func4(A):
out = A.T + A
mask = np.eye(out.shape[0],dtype=bool)
out[mask] = A[mask]
return out
def func5(A):
return np.where(np.eye(A.shape[0],dtype=bool),A,A.T+A)
def func6(A):
return np.where(np.triu(np.ones(A.shape[0],dtype=bool),1),A.T,A)
时间-
In [140]: # Input array
...: N = 5000
...: A = np.tril(np.random.randint(0,9,(N,N)))
...:
In [141]: %timeit func1(A)
...: %timeit func2(A)
...: %timeit func3(A)
...: %timeit func4(A)
...: %timeit func5(A)
...: %timeit func6(A)
...:
1 loops, best of 3: 617 ms per loop
1 loops, best of 3: 354 ms per loop
1 loops, best of 3: 354 ms per loop
1 loops, best of 3: 395 ms per loop
1 loops, best of 3: 597 ms per loop
1 loops, best of 3: 440 ms per loop
看起来像方法2& #3效率很高!
Looks like the approaches # 2 & #3 are pretty efficient!
这篇关于如何在NumPy中将三角矩阵转换为正方形?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!