稀疏地将一个稀疏矩阵的所有行附加到另一行

Question

I have a numpy matrix and want to append another matrix to that.

The two matrices have the shapes:

m1.shape = (2777, 5902)  m2.shape = (695, 5902)

I want to append m2 to m1 so that the new matrix is of shape:

m_new.shape = (3472, 5902)

When I use numpy.append or numpy.concatenate I just get a new array with the two matrix in it and the shape (2,1).

Any one of you have an Idea how to get one big matrix out of the two?

Additional info: both are sparse matrices.

EDIT: m1 looks like

(0, 1660)   0.444122811195
(0, 3562)   0.260868771714
(0, 4743)   0.288149437574
(0, 4985)   0.514889706991
(0, 5215)   0.272163636657
(0, 5721)   0.559006134727
(1, 555)    0.0992498400527
(1, 770)    0.133145289523
(1, 790)    0.0939044698233
(1, 1097)   0.259867567986
(1, 1285)   0.188836288168
(1, 1366)   0.24707459927
(1, 1499)   0.237997843516
(1, 1559)   0.120069347224
(1, 1701)   0.17660176488
(1, 1926)   0.185678520634
(1, 2177)   0.163066377369
(1, 2641)   0.079958199952
(1, 2937)   0.259867567986
(1, 3551)   0.198471489351
(1, 3562)   0.0926197593026
(1, 3593)   0.100537828805
(1, 4122)   0.198471489351
(1, 4538)   0.57162654484
(1, 4827)   0.105808609537

m2 looks like:

(0, 327)    0.0770581299315
  (0, 966)  0.309858753157
  (0, 1231) 0.286870892505
  (0, 1384) 0.281385698712
  (0, 1817) 0.204495931592
  (0, 2284) 0.182420951496
  (0, 2414) 0.114591086901
  (0, 2490) 0.261442040482
  (0, 3122) 0.321676138471
  (0, 3151) 0.286870892505
  (0, 4031) 0.172251612658
  (0, 5149) 0.25839783806
  (0, 5215) 0.125806303262
  (0, 5225) 0.336280781816
  (0, 5231) 0.135930403721
  (0, 5294) 0.145049459537
  (0, 5794) 0.20145172917
  (0, 5821) 0.224439589822
  (1, 327)  0.191031948626
  (1, 1171) 0.62081265022

Type of the matrices is:

<class 'scipy.sparse.csr.csr_matrix'> <class 'scipy.sparse.csr.csr_matrix'>

SOLVED:

m_new = scipy.sparse.vstack((m1, m2))

did the trick

Thanks for your help.

Answer 1

You can use numpy.vstack in your case (or numpy.hstack , when matrices shapes are (x,y) and (x,z))

Example:

a = np.zeros((3,7))
b = np.zeros((46,7))
c = np.vstack((a,b))
print c.shape
#(49,7)

Answer 2

You can use concatenate ( http://docs.scipy.org/doc/numpy-1.10.0/reference/generated/numpy.concatenate.html ):

m_new = np.concatenate((m1, m2))

For example,

a = np.zeros((5, 3))
b = np.zeros((4, 3))
print np.concatenate((a, b)).shape
# (9, 3)

That still works if you use matrices instead of arrays:

print np.concatenate((np.matrix(a), np.matrix(b))).shape
# (9, 3)