The scipy.sparse package provides a number of sparse matrix storage schemes like compressed sparse row/column, linked list or coordinate formats. Each of those is suitable for some applications and unsuitable for other ones. I will first introduce every format available and mention its strong points and weaknesses. I will also discuss some common difficulties, originating from the fact, that the sparse matrix objects are not subclasses of numpy.ndarray, namely fancy indexing issues.
Then I will dive into the related scipy.sparse.linalg module, that contains sparse eigenvalue, iterative and direct solvers, and show how to use those solvers.
Rough outline of the tutorial session:
- Sparse matrix formats in SciPy
- Introduction of the formats
- Common issues
- Sparse matrix solvers in SciPy
If you have an interesting sparse computing-related topic other then those mentioned above, post a comment, please.