These changes are in preparation for pygrank's third major release (0.3.0):
- The updated documentation can be found here: https://pygrank.readthedocs.io/en/latest/
- Backends now admit initialization parameters, such as preferred device and mode of operations. Additionally, GPU backends have a default mode of trying to use dense matrices to represent graphs, as long as those matrices reside in memory. This ensures that the fastest possible computations will be performed out-of-the-box.
- Backend bottlenecks have been optimized to require similar execution time as other fast computations.