Fast Cosine Similarity Search in Binary Space with Angular Multi-Index Hashing

Abstract

Given a large dataset of binary codes and a binary query point, we address how to efficiently find K codes in the dataset that yield the largest cosine similarities to the query. The straightforward answer to this problem is to compare the query with all items in the dataset, but this is practical only for small datasets. One potential solution to enhance the search time and achieve sublinear cost is to use a hash table populated with binary codes of the dataset and then look up the nearby buckets to the query to retrieve the nearest neighbors. However, if codes are compared in terms of cosine similarity rather than the Hamming distance, then the main issue is that the order of buckets to probe is not evident. To examine this issue, we first elaborate on the connection between the Hamming distance and the cosine similarity. Doing this allows us to systematically find the probing sequence in the hash table. However, solving the nearest neighbor search with a single table is only practical for short binary codes. To address this issue, we propose the angular multi-index hashing search algorithm which relies on building multiple hash tables on binary code substrings. The proposed search algorithm solves the exact angular K nearest neighbor problem in a time that is often orders of magnitude faster than the linear scan baseline and even approximation methods.Index Terms-Nearest neighbor search, binary codes, large-scale retrieval, cosine similarity ✦ • S. Eghbali is with the

The Improved Artificial Neural Network Based on Cosine Similarity in Facial Emotion Recognition

“…Even [8] showing cosine similarity measure has an accuracy of up to 0.99 in measuring semantic vectors that have been reduced to 8 bits. Moreover [9] also shows that CSM faster than the linear scan and approximation methods.…”

Improved Particle Swarm Optimization for Sea Surface Temperature Prediction

“…Cosine similarity [21] uses the cosine of the angles of two vectors in the vector space to measure the similarity between them. The closer the cosine value is to 1, the closer the angle is to 0, the more similar they are.…”

Deep Spherical Quantization for Image Search

“…For the local search, we again use alternative optimization technique. Given {b ij } j =t fixed, b it is updated by exhaustively checking all codewords of C j and finding the element that minimizes the objective function in (13). For the perturbation procedure of SLS, we randomly choose k codes by sampling from the uniform distribution U (1, m).…”

“…Binary-valued representation has several advantages, such as being compact to store and faster to compare, making it a suitable fit for large-scale nearest neighbor search. Moreover, for binary strings, one can achieve sublinear query time using hash tables [13,34] or tree-based indexing data structures [11,12]. Finding compact binary codes that better respect the given notion of similarity has been the topic of much work over the last two decades during which a rich set of hashing techniques has been proposed.…”