【摘要】 Minwise Hash极大似然估计子RMle综合考虑所有事件的发生概率,可以提高估计精度,但降低了估计的效率。连接位Minwise Hash估计子R_Minwise,c可以成倍减少比对次数,动态阈值过滤器能够进一步提高Minwise Hash算法和其变种算法的效率。结合连接位极大似然估计子和动态阈值过滤器,提出了连接位极大似然动态过滤算法R(T_Mle,c)。实验表明,R(T_Mle,c)具有精度和效率兼顾的特性,计算时间最少,并且在k>300的条件下,其准确度与RMle的近乎相等。更多还原

【Abstract】 Maximum likelihood estimator(R_Mle)can improve the average accuracy.Connected bit Minwise Hash(R_Minwise,c)can exponentially improve efficiency of similarity estimation.Dynamic threshold filter makes further improvement on efficiency.Combining RMinwise,cand dynamic threshold filter,the maximum likelihood dynamic filtering algorithm of Connected bit Minwise Hash was proposed.Experimental results demonstrate that R(T_Mle,c)can get second-best precision and efficiency,and is the most cost effective business in the estimator options(R_Mle,R_Mle,c,R_Minwise,c,R(T_Mle,c).更多还原