ICML2014-Nearest-Neighbors and Large-Scale Learning

- Track F – Nearest-Neighbors and Large-Scale Learning (Location: 305-2, Chair: Kilian Weinberger) [01]–10:50 Densifying One Permutation Hashing via Rotation for Fast Near Neighbor Search Anshumali Shrivastava; Ping Li 主要是为降低LSH Query的查询复杂度,扫看全文第一感觉有点类似于WWW2007 GoogleSimHash的查询思路,暂未认真阅读。 使用的数据集:MNIST,NEWS20,WEBSPAM [02]–11:10 Coding for Random Projections Ping Li; … 继续阅读

Reading Paper: ACL2014-Two-Stage Hashing for Fast Document Retrieval

基本思路:LSH + ITQ 的两阶段方法,思路比较清晰明了。 该Paper基于如下两个方法: LSH: A. Andoni and P. Indyk. 2008. Near-optimal hashing algorithms for approximate nearest neighbor in high dimensions. ommunications of the ACM, 51(1):117–122. ITQ: Y. Gong, S. Lazebnik, A. Gordo, and F. Perronnin. 2013. Iterative quantization: … 继续阅读