Community

linear algebra and fillna with Pandas

Using Pandas/Numpy, I took the inverse of a matrix as follows:

A = D.dot(B)   # 'D' is  234x462 and 'B' is 462x234
I = np.identity(234)   # Identity matrix that is 234x234
step_1 = I - A
squared = inv(step_1)

Before I did anything, I had to manually fill empty cells in the 'D' matrix with zeros.  After that, everything worked fine and I got the results I expected.  However,  I have a very large number of similar computations to perform and, as you can imagine, I don't want to manually fill empty cells in my dataframes every time I am performing a matrix inversion.  With that in mind, I used the fillna method as follows:

D = D.fillna(0)

This filled the empty cells with zeros and everything looked fine but when I try to do the matrix inversion (as above), I get an error message: "LinAlgError: Singular matrix."  Any ideas/suggestions?  Thanks!