Sparse matrix with non-zero default in scipy

Question

I would like to create a "sparse" matrix, where undefined entries are not zero, but some value. The matrix that I am trying to create calculates the result of certain types of interactions between elements i and j, but since there are much more records than interactions, this will certainly be very scarce - however, it would be useful to ensure a uniform preliminary belief in the result of these interactions (i.e. e. Laplace smoothing), which will make the matrix necessarily dense, albeit equally compressible. Thank!