中文摘要：

基于哈希算法的相似性检索以其高效性和实用性受到学术界的广泛关注。为了提升哈希学习算法的准确性和扩展性,提出了一种基于类标签的离散监督学习算法。首先,假设每个类标签都存在一个潜在的哈希码,并深入探索了类标签的关联性与其哈希码之间的关系,用以求解每个类的哈希码。然后,度量数据点的哈希码与类哈希码间的内积关系构建度量模型,同时采用非线性核函数建立量化模型。最后,在求解哈希码的过程中,采用了离散求解法以提升准确性。NUS-WIDE和CIFAR-10数据集的实验结果均表明,基于类标签的离散监督哈希算法是有效的。

英文摘要：

Similarity search based on hashing algorithm is widely concerned by the academic circles for its high efficiency and practicality. In order to improve the accuracy and scalability of the hashing algorithm,a discrete supervised learning algorithm was proposed based on class label. Firstly,under the assumption that for each class label there is a potential hash code,the relevance of the associated class labels and its hash code to achieving the hash code for each class is explored. Then,a model for the inner product of the relationship between the hash code and its label＇s hash code is established. Finally,discrete method to compute the model is proposed. The experimental results of NUS-WIDE and CIFAR-10 data sets show that the hashing algorithm based on the class label of the discrete supervised algorithm is effective.