PythonNewaxis vs for循环
我正在努力使我的程序更快。 我有一个矩阵和一个向量:PythonNewaxis vs for循环,python,loops,numpy,Python,Loops,Numpy,我正在努力使我的程序更快。 我有一个矩阵和一个向量: GDES = N.array([[1,2,3,4,5], [6,7,8,9,10], [11,12,13,14,15], [16,17,18,19,20], [21,22,23,24,25]]) Ene=N.array([1,2,3,4,5]) NN=len(GDES); 我定义了一个矩阵乘法函数: def Gl(n,np,k,q): matrix = GDES[k,np]*GDES[k,n]*GDES[q,np]*GDES[q,n
GDES = N.array([[1,2,3,4,5],
[6,7,8,9,10],
[11,12,13,14,15],
[16,17,18,19,20],
[21,22,23,24,25]])
Ene=N.array([1,2,3,4,5])
NN=len(GDES);
我定义了一个矩阵乘法函数:
def Gl(n,np,k,q):
matrix = GDES[k,np]*GDES[k,n]*GDES[q,np]*GDES[q,n]
return matrix
我在计算中做了一个for循环:
SIl = N.zeros((NN,NN),N.float)
for n in xrange(NN):
for np in xrange(NN):
SumJ = N.sum(N.sum(Gl(n,np,k,q) for q in xrange(NN)) for k in xrange(NN))
SIl[n,np]=SumJ
print 'SIl:',SIl
输出:
SIl: [[ 731025. 828100. 931225. 1040400. 1155625.]
[ 828100. 940900. 1060900. 1188100. 1322500.]
[ 931225. 1060900. 1199025. 1345600. 1500625.]
[ 1040400. 1188100. 1345600. 1512900. 1690000.]
[ 1155625. 1322500. 1500625. 1690000. 1890625.]]
MGI: [[ 825 3900 9225 16800 26625]
[ 31200 92400 169600 262800 372000]
[ 146575 413400 722475 1073800 1467375]
[ 403200 1116900 1911600 2787300 3744000]
[ 857325 2352900 3980725 5740800 7633125]]
我想使用newaxis
来加快速度:
def G():
Mknp = GDES[:, :, N.newaxis, N.newaxis]
Mkn = GDES[:, N.newaxis, :, N.newaxis]
Mqnp = GDES[:, N.newaxis, N.newaxis, :]
Mqn = GDES[N.newaxis, :, :, N.newaxis]
matrix=Mknp*Mkn*Mqnp*Mqn
return matrix
tmp = G()
MGI = N.sum(N.sum(tmp,axis=3), axis=2)
MGI = N.reshape(MGI,(NN,NN))
print 'MGI:', MGI
输出:
SIl: [[ 731025. 828100. 931225. 1040400. 1155625.]
[ 828100. 940900. 1060900. 1188100. 1322500.]
[ 931225. 1060900. 1199025. 1345600. 1500625.]
[ 1040400. 1188100. 1345600. 1512900. 1690000.]
[ 1155625. 1322500. 1500625. 1690000. 1890625.]]
MGI: [[ 825 3900 9225 16800 26625]
[ 31200 92400 169600 262800 372000]
[ 146575 413400 722475 1073800 1467375]
[ 403200 1116900 1911600 2787300 3744000]
[ 857325 2352900 3980725 5740800 7633125]]
你知道我怎样才能得到正确答案吗?你的问题非常适合: 对于您的特定情况,这种其他语法可能更容易理解:
>>> np.einsum(GDES, [2,1], GDES, [2,0], GDES, [3,1], GDES, [3,0], [0,1])
array([[ 731025, 828100, 931225, 1040400, 1155625],
[ 828100, 940900, 1060900, 1188100, 1322500],
[ 931225, 1060900, 1199025, 1345600, 1500625],
[1040400, 1188100, 1345600, 1512900, 1690000],
[1155625, 1322500, 1500625, 1690000, 1890625]])
您的问题非常适合: 对于您的特定情况,这种其他语法可能更容易理解:
>>> np.einsum(GDES, [2,1], GDES, [2,0], GDES, [3,1], GDES, [3,0], [0,1])
array([[ 731025, 828100, 931225, 1040400, 1155625],
[ 828100, 940900, 1060900, 1188100, 1322500],
[ 931225, 1060900, 1199025, 1345600, 1500625],
[1040400, 1188100, 1345600, 1512900, 1690000],
[1155625, 1322500, 1500625, 1690000, 1890625]])