Biblio

A fundamental recurring task in many machine learning applications is the search for the Nearest Neighbor in high dimensional metric spaces. Towards answering queries in large scale problems, state-of-the-art methods employ Approximate Nearest Neighbors (ANN) search, a search that returns the nearest neighbor with high probability, as well as techniques that compress the dataset. Product-Quantization (PQ) based ANN search methods have demonstrated state-of-the-art performance in several problems, including classification, regression and information retrieval. The dataset is encoded into a Cartesian product of multiple low-dimensional codebooks, enabling faster search and higher compression. Being intrinsically parallel, PQ-based ANN search approaches are amendable for hardware acceleration. This paper proposes a novel Hierarchical PQ (HPQ) based ANN search method as well as an FPGA-tailored architecture for its implementation that outperforms current state of the art systems. HPQ gradually refines the search space, reducing the number of data compares and enabling a pipelined search. The mapping of the architecture on a Stratix 10 FPGA device demonstrates over ×250 speedups over current state-of-the-art systems, opening the space for addressing larger datasets and/or improving the query times of current systems.

Efficient and secure search on encrypted data is an important problem in computer science. Users having large amount of data or information in multiple documents face problems with their storage and security. Cloud services have also become popular due to reduction in cost of storage and flexibility of use. But there is risk of data loss, misuse and theft. Reliability and security of data stored in the cloud is a matter of concern, specifically for critical applications and ones for which security and privacy of the data is important. Cryptographic techniques provide solutions for preserving the confidentiality of data but make the data unusable for many applications. In this paper we report a novel approach to securely store the data on a remote location and perform search in constant time without the need for decryption of documents. We use bloom filters to perform simple as well advanced search operations like case sensitive search, sentence search and approximate search.