大数据跨源哈希学习的理论和应用研究

Since hashing learning representing high-dimensional data with low-dimensional binary codes can reduce the store cost, the retrieval complexity, and avoids the issue of curse of dimensionality, it has quickly been used in all kinds of domains of big data, such as image retrieval, recommendation system and social network. Conventional hashing learning methods consist of data-independent hashing and data-dependent hashing, where data-independent hashing has with stronger theoretical demonstration but less retrieval performance than data-dependent hashing. Based on the heterogeneity and diversity of big data, this proposal exploits the rule of information retrieval on big data via studying the theory and real applications of cross-media hashing methods. More specifically, under the assumption of the basic theory of data-dependent hashing and information theory, this proposal expects to discover the theoretical models of data-dependent hashing and to design efficient cross-media hashing methods on heterogeneous big data, such as non-structural data, high-dimensional data, multi-level data, followed by constructing the computer-aided diagnosis models on neuroimaging data of Alzheimer's disease, for alleviating the status of social / family problems in our country due to the increasing phenomenon of the issue of low-birth rate and ageing.

哈希学习将高维数据表示成低维二进制能降低大数据的存储开销和检索复杂度，同时也能处理维度灾难问题，从而快速地在图像检索、推荐系统和社交网络等大数据领域得到广泛的应用。常见的哈希学习方法有数据依赖哈希和数据独立哈希两种，其中数据独立哈希有理论依据但检索效果不如数据依赖哈希。基于在跨源大数据的异质性和多样性等特点，本项目拟通过研究大数据检索中跨源哈希学习的理论和应用探索大数据信息检索规律。具体地说，本项目拟以数据独立哈希学习方法的基础理论和和信息编码理论为工具，从实际应用出发，揭示数据依赖哈希学习理论模型和研究各种异质（例如非结构化、高维、多粒度等）跨源哈希学习方法和技术，并在健康大数据应用中建立基于神经影像检查结果的阿尔茨海默症计算机辅助诊断检索模型，以缓解现阶段我国日益加重的老年少子化现象而导致社会/家庭等社会问题。

哈希学习将高维数据表示成低维二进制能降低大数据的存储开销和检索复杂度，同时也能处理维度灾难问题，从而快速地在图像检索、推荐系统和社交网络等大数据领域得到广泛的应用。常见的哈希学习方法有数据依赖哈希和数据独立哈希两种，其中数据独立哈希有理论依据但检索效果不如数据依赖哈希。基于在跨源大数据的异质性和多样性等特点，本项目拟通过研究大数据检索中跨源哈希学习的理论和应用探索大数据信息检索规律。具体地说，本项目拟以数据独立哈希学习方法的基础理论和和信息编码理论为工具，从实际应用出发，揭示数据依赖哈希学习理论模型和研究各种异质（例如非结构化、高维、多粒度等）跨源哈希学习方法和技术，并在健康大数据应用中建立基于神经影像检查结果的阿尔茨海默症计算机辅助诊断检索模型，以缓解现阶段我国日益加重的老年少子化现象而导致社会/家庭等社会问题。