Block tridiagonal matrix python

Question

I would like to create a block tridiagonal matrix starting from three numpy.ndarray. Is there any (direct) way to do that in python?

Thank you in advance!

Ch

Answer 1

For better or worse, all the other answers seem to answer about tridiagonal matrices and not block tridiagonal matrices.

I don't think there is native support for tridiagonal matrices, so I wrote my own code. I had zeros on the main diagonal and my matrix was symmetric.

Here is my code.

n1 = 784
n2 = 256
n3 = 128
n4 = 10
M1 = np.ones((n1,n2))
M2 = np.ones((n2,n3))
M3 = np.ones((n3, n4))

def blockTri(Ms):
    #Takes in a list of matrices (not square) and returns a tridiagonal block matrix with zeros on the diagonal
    count = 0
    idx = []
    for M in Ms:
        #print(M.shape)
        count += M.shape[0]
        idx.append(count)
    count += Ms[-1].shape[-1]
    mat = np.zeros((count,count))
    count = 0
    for i, M in enumerate(Ms):
        mat[count:count+M.shape[0],idx[i]:idx[i]+M.shape[1]] = M
        count = count + M.shape[0]
    mat = mat + mat.T    
    return mat

M = blockTri([M1, M2, M3])

Hopefully this can help future people looking for block tridiagonal matrices.