Algorithm Specification
The complete formal specification is available in the repository.
See spec/SPEC.md for:
- Formal algorithm definition
- Pseudocode for all operations
- Mathematical foundations
- Implementation requirements
- Test vector format
Key Sections
- Overview — Algorithm purpose and design goals
- Normalization — L2 norm calculation
- SplitMix64 PRNG — Deterministic random generation
- Hyperplane Generation — Seed formula and sign extraction
- SimHash — Bit packing and hex encoding
- Hamming Distance — Bit difference calculation
- Cosine Estimation — Similarity from Hamming distance
- SHA256 Canonical — Deterministic embedding hashing
- Signature Format — String representation
- Banding — LSH indexing strategy
- CM-LSH — Confidence Matrix extension
- Cross-Language — Validation requirements
- Test Vectors — Canonical test cases