Random Projection Location Sensitive Hash in Go

Random Projection LSH is a way to create a binary hash from a vector, with closer vectors having a higher chance of having the same hash.

Steps

• create a random vector with the length of the hash
• calculate the dot product of the to-be-encoded hash with the random vector
• with output: All positive values become 1, all negative values become 0

Install

go get github.com/jurschreuder/rplsh

New random hash table

To hash vectors of []float64 with length 128.
Create a hashes of 16 bits.

ht := NewHashTable(128, 16)

Create a []byte hash

vecs := make([]float64, 128)

bs := ht.Hash(vecs)

Create a uint hash

Supports: uint8, uint16, uint32, uint64.

vecs := make([]float64, 128)

bs := ht.HashUint16(vecs)

Save hash table for consistent results

err := ht.Save("save_file.js")

Load hash table from save file

hs, err := ht.Load("save_file.js")