Appendix 602 A Design of LSH Functions: 603 In practice, we could realize d (x, y)

As shown in Definition A.1, the SRP hash is an LSH function built upon random Gaussian projections, The SRP hash is usually designed for dense vectors. Following Definition A.2, MinHash serves as a powerful tool for Jaccard similarity estimation [ In this paper, we take a kernel view of the collision probability of LSH (see Definition 3.2). In this section, we introduce the client selection algorithm with our one-pass sketch. To formally prove the Theorems in the paper. Next, we introduce the formal statements in the paper as below.