Add to_numpy_array or to_numpy_matrix
#342
Comments
|
There is an |
|
Thank for your information. It would be great if it is mentioned in https://qiskit.org/documentation/retworkx/networkx.html |
mtreinish
added a commit
to mtreinish/retworkx
that referenced
this issue
May 28, 2021
This commit adds documentation to the retworkx for networkx users guide on the differences between networkx and retworkx when it comes to matrix generation. Right now retworkx only has an `adjacency_matrix()` function which is equivalent to networkx's `to_numpy_array()` and returns a numpy array of the adjacency matrix. This is different from networkx's `adjacency_matrix()` function which returns a scipy sparse array (and is just an alias for networkx's `to_scipy_sparse_matrix()`). Documenting this gap is important because it's easy to get confused by these API differences. This difference is fundamentally because we can effeciently create numpy arrays in retworkx by just pass a pointer to numpy with a 2d array that was constructed natively in rust (using rust-numpy) while other matrix formats (such as scipy sparse matrices) don't expose a C api so to create those objects we'd have to go through Python. Eventually we might expand the API to include pure python functions to do additional conversions (although I worry about increasing the external dependency list) but numpy arrays will still be the default because of the inherent performance benefits. Fixes Qiskit#342
mtreinish
added a commit
that referenced
this issue
May 28, 2021
…#343) This commit adds documentation to the retworkx for networkx users guide on the differences between networkx and retworkx when it comes to matrix generation. Right now retworkx only has an `adjacency_matrix()` function which is equivalent to networkx's `to_numpy_array()` and returns a numpy array of the adjacency matrix. This is different from networkx's `adjacency_matrix()` function which returns a scipy sparse array (and is just an alias for networkx's `to_scipy_sparse_matrix()`). Documenting this gap is important because it's easy to get confused by these API differences. This difference is fundamentally because we can effeciently create numpy arrays in retworkx by just pass a pointer to numpy with a 2d array that was constructed natively in rust (using rust-numpy) while other matrix formats (such as scipy sparse matrices) don't expose a C api so to create those objects we'd have to go through Python. Eventually we might expand the API to include pure python functions to do additional conversions (although I worry about increasing the external dependency list) but numpy arrays will still be the default because of the inherent performance benefits. Fixes #342
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
What is the expected enhancement?
It would be great if retworkx supports
to_numpy_arrayorto_numpy_matrix.It is used in a tutorial of qiskit-optimization.
https://github.com/Qiskit/qiskit-optimization/blob/main/docs/tutorials/06_examples_max_cut_and_tsp.ipynb
It is not urgent because we can implement a utility function in Python.
Related to qiskit-community/qiskit-optimization#150
The text was updated successfully, but these errors were encountered: