From this mwe:
a=np.zeros((5,5))
b=np.zeros((2,2))
a=np.matrix(a)
b=np.matrix(b)
b[0,0]=4
b[1,1]=9
b[0,1]=7
indice=[2,3]
# 1
c=a[indice,:][:,indice]
c=b
print c
# 2
a[indice,:][:,indice]=b
print a[indice,:][:,indice]
I get:
>>> c
matrix([[ 4., 7.],
[ 0., 9.]])
and:
>>> a[indice,:][:,indice]
matrix([[ 0., 0.],
[ 0., 0.]])
I do not understand why the values of a remain zeroes. If a similar operation is done in two steps, everything works fine:
>>> for k in range(len(indice)):
... a[indice[k],indice]=b[k,:]
I obtain:
>>> a
matrix([[ 0., 0., 0., 0., 0.],
[ 0., 4., 0., 7., 0.],
[ 0., 0., 0., 0., 0.],
[ 0., 0., 0., 9., 0.],
[ 0., 0., 0., 0., 0.]])