Hamming Distance Metric Learning
نویسندگان
چکیده
Motivated by large-scale multimedia applications we propose to learn mappings from high-dimensional data to binary codes that preserve semantic similarity. Binary codes are well suited to large-scale applications as they are storage efficient and permit exact sub-linear kNN search. The framework is applicable to broad families of mappings, and uses a flexible form of triplet ranking loss. We overcome discontinuous optimization of the discrete mappings by minimizing a piecewise-smooth upper bound on empirical loss, inspired by latent structural SVMs. We develop a new loss-augmented inference algorithm that is quadratic in the code length. We show strong retrieval performance on CIFAR-10 and MNIST, with promising classification results using no more than kNN on the binary codes.
منابع مشابه
Improving Unconstrained Iris Recognition Performance via Domain Adaptation Metric Learning Method
To improve unconstrained iris recognition system performance in different environments, a performance improvement method of unconstrained iris recognition based on domain adaptation metric learning is proposed. A kernel matrix is calculated as the solution of domain adaptation metric learning. The known Hamming distance computing by intra-class and inter-class is used as the optimization learni...
متن کاملOn deep-holes of Gabidulin codes
In this paper, we determine a class of deep holes for Gabidulin codes with both rank metric and Hamming metric. Moreover, we give a necessary and sufficient condition for deciding whether a word is not a deep hole for Gabidulin codes, by which we study the error distance of two special classes of words to certain Gabidulin codes.
متن کاملOn the minimality of Hamming compatible metrics
A Hamming compatible metric is an integer-valued metric on the words of a finite alphabet which agrees with the usual Hamming distance for words of equal length. We define a new Hamming compatible metric, compute the cardinality of a sphere with respect to this metric, and show this metric is minimal in the class of all “well-behaved” Hamming compatible metrics.
متن کاملThe Canonical Metric for Vector Quantization
To measure the quality of a set of vector quantization points a means of measuring the distance between two points is required. Common metrics such as the Hamming and Euclidean metrics, while mathematically simple, are inappropriate for comparing speech signals or images. In this paper it is argued that there often exists a natural environment of functions to the quantizationprocess (for exampl...
متن کاملWeighted Hamming Metric Structures
A weighted Hamming metric is introduced in [4] and it showed that the binary generalized Goppa code is a perfect code in some weighted Hamming metric. In this paper, we study the weight structures which admit the binary Hamming code and the extended binary Hamming code to be perfect codes in the weighted Hamming metric. And, we also give some structures of a 2-perfect code and how to construct ...
متن کاملImage Retrieval and Classification Using Local Distance Functions
(x − x)A(x − x) Mahalanobis distance: Previous work on learning metrics has focused on learning a single distance metric for all instances. One of our primary contributions is to learn a distance function for every training image. Most visual categorization approaches make use of machine learning after computing distances between images (e.g. SVM with pyramid kernel). We want to learn how to co...
متن کامل