Geek asks and answers

Adding a very repeating matrix to sparse in numpy / scipy?

I am trying to implement a function in NumPy / Scipy to calculate the Jensen-Shannon discrepancy between one (training) vector and a large number of other (observational) vectors. Observation vectors are stored in a very large (500 000x65536) Scipy sparse matrix (a dense matrix will not fit into memory).

As part of the algorithm, I need to calculate T + O i for each observation vector O iwhere T is the training vector. I could not find a way to do this using the usual NumPy broadcast rules, since sparse matrices do not seem to support them (if T remains as a dense array, Skipy first tries to cut the sparse matrix tightly, which if I make T into a sparse matrix, T + O i fails because the forms are incompatible).

Currently, I am taking an extremely inefficient step of dividing the training vector into a sparse matrix 500 000x65536:

training = sp.csr_matrix(training.astype(np.float32))
tindptr = np.arange(0, len(training.indices)*observations.shape[0]+1, len(training.indices), dtype=np.int32)
tindices = np.tile(training.indices, observations.shape[0])
tdata = np.tile(training.data, observations.shape[0])
mtraining = sp.csr_matrix((tdata, tindices, tindptr), shape=observations.shape)

But it takes up a huge amount of memory (about 6 GB) when it stores only 1,500 "real" elements. It also creates quite slowly.

I tried to get smart using stride_tricks to make insplr CSR matrices and data members not use extra memory for repeated data.

training = sp.csr_matrix(training)
mtraining = sp.csr_matrix(observations.shape,dtype=np.int32)
tdata = training.data
vdata = np.lib.stride_tricks.as_strided(tdata, (mtraining.shape[0], tdata.size), (0, tdata.itemsize))
indices = training.indices
vindices = np.lib.stride_tricks.as_strided(indices, (mtraining.shape[0], indices.size), (0, indices.itemsize))
mtraining.indptr = np.arange(0, len(indices)*mtraining.shape[0]+1, len(indices), dtype=np.int32)
mtraining.data = vdata
mtraining.indices = vindices

, strided views mtraining.data mtraining.indices ( ). , .flat, , , (, dtype), flatten() .

?

+5
1

, , , , . , scipy :

>>> a = sps.csr_matrix([1,2,3,4])
>>> a.data
array([1, 2, 3, 4])
>>> a.indices
array([0, 1, 2, 3])
>>> a.indptr
array([0, 4])

>>> b = sps.csr_matrix((np.array([1, 2, 3, 4, 5]),
...                     np.array([0, 1, 2, 3, 0]),
...                     np.array([0, 5])), shape=(1, 4))
>>> b
<1x4 sparse matrix of type '<type 'numpy.int32'>'
    with 5 stored elements in Compressed Sparse Row format>
>>> b.todense()
matrix([[6, 2, 3, 4]])

, , : , .

, :

def csr_add_sparse_vec(sps_mat, sps_vec) :
    """Adds a sparse vector to every row of a sparse matrix"""
    # No checks done, but both arguments should be sparse matrices in CSR
    # format, both should have the same number of columns, and the vector
    # should be a vector and have only one row.

    rows, cols = sps_mat.shape
    nnz_vec = len(sps_vec.data)
    nnz_per_row = np.diff(sps_mat.indptr)
    longest_row = np.max(nnz_per_row)

    old_data = np.zeros((rows * longest_row,), dtype=sps_mat.data.dtype)
    old_cols = np.zeros((rows * longest_row,), dtype=sps_mat.indices.dtype)

    data_idx = np.arange(longest_row) < nnz_per_row[:, None]
    data_idx = data_idx.reshape(-1)
    old_data[data_idx] = sps_mat.data
    old_cols[data_idx] = sps_mat.indices
    old_data = old_data.reshape(rows, -1)
    old_cols = old_cols.reshape(rows, -1)

    new_data = np.zeros((rows, longest_row + nnz_vec,),
                        dtype=sps_mat.data.dtype)
    new_data[:, :longest_row] = old_data
    del old_data
    new_cols = np.zeros((rows, longest_row + nnz_vec,),
                        dtype=sps_mat.indices.dtype)
    new_cols[:, :longest_row] = old_cols
    del old_cols
    new_data[:, longest_row:] = sps_vec.data
    new_cols[:, longest_row:] = sps_vec.indices
    new_data = new_data.reshape(-1)
    new_cols = new_cols.reshape(-1)
    new_pointer = np.arange(0, (rows + 1) * (longest_row + nnz_vec),
                            longest_row + nnz_vec)

    ret = sps.csr_matrix((new_data, new_cols, new_pointer),
                         shape=sps_mat.shape)
    ret.eliminate_zeros()

    return ret

, , 10 000 1 .:

In [2]: a
Out[2]: 
<10000x65536 sparse matrix of type '<type 'numpy.float64'>'
    with 15000000 stored elements in Compressed Sparse Row format>

In [3]: b
Out[3]: 
<1x65536 sparse matrix of type '<type 'numpy.float64'>'
    with 1500 stored elements in Compressed Sparse Row format>

In [4]: csr_add_sparse_vec(a, b)
Out[4]: 
<10000x65536 sparse matrix of type '<type 'numpy.float64'>'
    with 30000000 stored elements in Compressed Sparse Row format>

In [5]: %timeit csr_add_sparse_vec(a, b)
1 loops, best of 3: 956 ms per loop

EDIT ,

def csr_add_sparse_vec(sps_mat, sps_vec) :
    """Adds a sparse vector to every row of a sparse matrix"""
    # No checks done, but both arguments should be sparse matrices in CSR
    # format, both should have the same number of columns, and the vector
    # should be a vector and have only one row.

    rows, cols = sps_mat.shape

    new_data = sps_mat.data
    new_pointer = sps_mat.indptr.copy()
    new_cols = sps_mat.indices

    aux_idx = np.arange(rows + 1)

    for value, col in itertools.izip(sps_vec.data, sps_vec.indices) :
        new_data = np.insert(new_data, new_pointer[1:], [value] * rows)
        new_cols = np.insert(new_cols, new_pointer[1:], [col] * rows)
        new_pointer += aux_idx

    return sps.csr_matrix((new_data, new_cols, new_pointer),
                          shape=sps_mat.shape)
+3

More articles:


All Articles