LSH(局部敏感哈希)算法

 

 

import org.apache.spark.mllib.linalg.SparseVector
import scala.util.Random

/**
  * @param hashNum 签名向量的维度， hash函数的个数
  */
class MinHash(hashNum: Int) extends Serializable {
    
    val HASH_PRIME=2038074743
    val rand = new Random()
    
    /**
    * n个随机哈希函数的参数配置
    */
    val randCoefs: Array[(Int, Int)] = Array.fill(hashNum) {
        (1 + rand.nextInt(HASH_PRIME - 1), rand.nextInt(HASH_PRIME - 1))
    }

    def generateSignature(vector: SparseVector): Array[Int] = {
        val indexes = vector.indices
        val signatureVector = randCoefs.map { 
            case (a, b) => 
            indexes.map(index => ((1 + index) * a + b) % HASH_PRIME).min
        }
        signatureVector
    }
}