bigKNN 0.3.0
- Added execution plans with memory-budget-based block sizing and
thread hints.
- Added serializable prepared references with
knn_load_prepared() and
knn_validate_prepared().
- Added exact radius-graph helpers, streamed directed kNN graph jobs,
checkpointed radius streaming, and resumable checkpoints.
- Added result coercion helpers plus exact recall summaries and
candidate reranking utilities.
bigKNN 0.2.0
- Added prepared exact references with
knn_prepare_bigmatrix(),
knn_search_prepared(), and
knn_search_stream_prepared() for repeated search
workloads.
- Added exact radius-search APIs with in-memory and streamed outputs,
including
count_within_radius_bigmatrix(),
radius_bigmatrix(), and
radius_stream_bigmatrix().
- Expanded metric support from Euclidean distance to
"sqeuclidean" and "cosine".
- Added exact graph builders for directed kNN, mutual kNN, and
shared-nearest-neighbour graphs.
bigKNN 0.1.0
- Initial release with exact Euclidean
k-nearest-neighbour search on
bigmemory::big.matrix references.
- Added self-search and external-query search against
big.matrix and dense matrix inputs.
- Added both in-memory and streamed neighbour outputs through
knn_bigmatrix() and
knn_stream_bigmatrix().
- Established package documentation, compiled backend registration,
and the first regression tests.