v1.1.0
This release has some performance improvements, both for accuracy and runtime.
- better graph indexing:
- extend graph windows on a traversal if the next window sketch will be identical
- merge graph windows for different traversals from the same start node if sketches are identical
- better alignment:
- use multi hash version of ntHash for KHV minhashing
- improve exact hierarchical alignment process
- better code:
- swaps some data structures over for better resource usage (some channels replacing constantly changing slices) and reduce the amount of pointers being passed between functions
** by better - these changes improve runtime, memory usage and accuracy reported by my current tests... Let me know if these changes are working out for you!