README
¶
Locality Sensitive Hashing (LSH)
This folder contains an educational implementation of LSH using random hyperplanes (sign hashing) to bucketize vectors for approximate nearest neighbor search.
Key Points
- Generates
numBitsrandom hyperplanes to encode vectors into binary signatures - Buckets are addressed by the signature; nearby buckets are explored by flipping bits within a radius
- Query expands radius until it gathers at least k candidates, then ranks by cosine distance
API (Unified)
- NewLSHIndex(d, numBits, seed) *LSHIndex
- Insert(vector []float64) error
- Search(query []float64, k int) ([]SearchResult, error)
- GetStats() map[string]interface{}
- GetNode(id int) interface{}
Usage
See examples/lsh_example.go and the repository root README for a quick start.
Notes
- Distance metric for ranking is cosine distance (1 - cosine similarity)
Documentation
¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func NewLshUtil ¶
Types ¶
type LSHIndex ¶
type LSHIndex struct {
// contains filtered or unexported fields
}
LSHIndex provides a unified interface over lshUtil
func NewLSHIndex ¶
NewLSHIndex creates a new LSH index with given dimension and bit count seed controls the random hyperplanes generation
type SearchResult ¶
SearchResult represents a search result with distance
Source Files
Click to show internal directories.
Click to hide internal directories.