All notable changes to this project will be documented in this file.
- Fixed random walk issue where node has no neighbors (rw is terminated)
- Fixed handling when a python error is thrown (into causes hangs).
- Fixed reservoir sampling iteration counter (to start at k instead of 0)
- Updated randomwalk to use a nonexistent inital prev not to avoid initial bias toward bfs
- Added changelog and changelog generation
- Added neighbor sampling example code
- Added an example for using random_walk function binding
- Added heterogenous graph neighborhood sampling
- Added sampler and filter specification support for python functions (for neighbor sampling)
- Added neighborhood sampling filter support. Added temporal and identity filters along with tests
- Added swappable node samplers for neighbor sampling to avoid duplicate code
- Added edge attribute type for sparse graphs
- Added global random state and random state reseeding
- Added python bindings for random_walk, neighbor_sampling and data conversion (csr, csc)
- Added weighted homogenous neighborhood sampling and some unit tests
- Added implementation for weighted sampling
- Added implementation for minheap on mutable slice
- Added node2vec random walk function
- Added csc edge based cumulative sum of attributes
- Added csc edge sorting by weights
- Moved interfaces to a separate module
- Added conversion of edge_index data to csc
- Added list or range sampling methods
- Added csc graph structure
- Added demo to test neighbourhood sampling
- Added replica of graph neighbourhood sampling
- Initial commit
- Moved error conversion to pyo3 into extension feature
- Swapped meaning of rows and cols in edge index data (now row (src) and col (dst))
- Moved neighbor sampling into its own namespace
- Code style fix
- Updated SparseGraph definition to be generic over CSR and CSC formats
- Added datasets for testing
- Temporarily removed CI building