Expand description
Cut-Aware HNSW: Dynamic Min-Cut Integration with Vector Search
This module bridges dynamic minimum cut tracking with HNSW vector search, enabling coherence-aware navigation that respects graph boundaries.
§Overview
Traditional HNSW blindly follows similarity edges during search. This module adds “coherence gates” - weak cuts in the graph that represent semantic boundaries. When searching, we can optionally halt expansion at these boundaries to stay within coherent regions.
§Key Concepts
- Cut Value: The total weight of edges crossing a partition
- Coherence Boundary: A weak cut indicating semantic separation
- Gated Search: Search that respects coherence boundaries
- Coherence Zone: A region of the graph with strong internal connections
§References
- Stoer-Wagner algorithm for global min-cut
- Euler Tour Trees for dynamic connectivity
- HNSW for approximate nearest neighbor search
Structs§
- Coherence
Zone - A coherent region in the graph
- CutAware
Config - Configuration for cut-aware HNSW
- CutAwareHNSW
- Extended HNSW that respects coherence boundaries
- CutAware
Metrics - Performance metrics for cut-aware operations
- Dynamic
CutWatcher - Dynamic cut watcher using incremental min-cut updates
- Edge
Update - Edge update operation
- Layer
CutStats - Cut statistics per layer
- Search
Result - Search result with coherence information
- Update
Stats - Statistics from batch update
Enums§
- Update
Kind - Type of edge update