论文-《Towards Optimal Binary Code Learning via Ordinal Embedding》

版权声明:本文为博主原创文章,未经博主允许不得转载。 https://blog.csdn.net/oniy_/article/details/81976321

Introduction

  1. 以前的哈希算法仅仅利用原始空间数据的相似度信息,本文的算法尝试将原始空间的排序关系映射到的低纬度的汉明空间中;
  2. approach constructs a directed, unweighted ordinal graph to capture the original relation among data points, upon which
    we learn hash functions via a novel stochastic gradient decreasing (SGD) algorithm

Preliminaries


OEH

构造一个有向的无权中的图G,图中每个点表示2个数据之间的不相似度,邮有向边表示4个数据点之间的排序关系。论文【construct a directed unweighted graph G = (V; E) = [n^4], where each node vij ⊆ V represents the dissimilarity degree between the i-th and the j-th sample, and each directed edge e(i;j;k;l)2C = (vij ! vkl) ⊆ E represents δij < δkl】

哈希函数:

待续...

猜你喜欢

转载自blog.csdn.net/oniy_/article/details/81976321