--- tags: [meeting, notes] --- *Attendees*: Isaac Virshup (@ivirshup), Jim Kitchen (@jim22k), Simon Cross (@hodgestar), Eric Gigure (@Ericgig) # Sparse interest meetup * Jim – at anaconda, working on python-graphblas * Eric– Sherbrooke QuTiP * Rewrote sparse operations because scipy was too slow * Simon Cross – QuTiP RIKEN * Implemented cython based sparse arrays within qutip * Problems with scipy sparse: * Sparse addition doing uneccesary copies * Inititalization and complex types * There are lots of checks making things slow * Downcasting indices/ indptr * Linear algebra and ODE use cases * Jim graph blas/ suite sparse * "sparse values" that aren't zero * suite sparse * Masked arrays * COO, CSR, and * MLIR graph blas * python-graphblas * Not Jit * Matrices must be 64 bit * On disk formats (degree of specialization) * Out of core needs? * Not doing a ton of this yet, but broad interest * Dask Graph blas from Eric Welch ### Links to projects * [AnnData](https://anndata.readthedocs.io/en/latest/) (ivirshup) * [qutip](https://qutip.org) * [python-graphblas](https://python-graphblas.readthedocs.io/en/latest/) ## Follow up * We will follow up on sparse on the discourse/ * TODO: Ask scientific python about where to discuss