Vector transformation for indexing, similarity search and classification

Abstract

A feature vector is encoded into a sparse binary vector. The feature vector is retrieved, for example from storage or a feature vector generator. The feature vector represents a media object or other data object. One or more permutations are generated, the dimensionality of the generated permutations equivalent to the dimensionality of the feature vector. The permutations may be generated randomly or formulaically. The feature vector is permuted with the one or more permutations, creating one or more permuted feature vectors. The permuted feature vectors are truncated according to a selected window size. The indexes representing the maximum values of the permuted feature vectors are identified and encoded using one -hot encoding, producing one or more sparse binary vectors. The sparse binary vectors may be concatenated into a single sparse binary vector and stored.; The sparse binary vector may be used in the similarity search, indexing or categorization of media objects.

Description

[0001] Cross References to Applications [0002] This application claims priority to Provisional Application No. 61 / 412,711, filed November 11, 2010, which is hereby incorporated by reference. technical field [0003] The present disclosure relates generally to the fields of data indexing, similarity searching and classification, and more specifically to manipulating high-dimensional vector space data. Background technique [0004] Vectors are often used to represent the feature space of various phenomena. For example, vectors are used to characterize images, videos, audio clips, and other media. It should be noted that the utility of vector space operations is not limited to digital media, but can also be applied to other data, to physical objects, or to any other entity capable of feature representation. In media space, features include color distribution (e.g. using 4x4 pixel hue and saturation histograms), mean and variance of color density across color channels, colo...

AI Technical Summary

Problems solved by technology

The magnitude of the mixed noise in the distance determination may exceed the magnitude of the change in the distance determination due to changes to a single vector dimension in a high-dimensional vector space

This is problematic in instances where it is desirable to measure the change in distance between vectors caused by a change in a small number of elements in the vector

Images