项目名称: 稀疏学习模型的核心理论和技术研究
项目编号: No.61263035
项目类型: 地区科学基金项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 朱晓峰
作者单位: 广西师范大学
项目金额: 44万元
中文摘要: 随着网络技术的快速发展,海量数据处理已经成为一个国际性需求的问题。开发有效处理海量数据的方法或者工具能影响到我们的日常生活,例如,很好地处理相机和手机接触到的海量信息,有利于设计和生产出速度更快和处理能力更强数码相机或者手机。处理海量数据问题的关键是准确的表述数据和有效的知识挖掘。本项目拟研究稀疏编码学习理论及其应用方面,对现有的稀疏编码方法进行扩展,建立准确表达现实中海量数据的方法和挖掘隐藏知识的理论模型。本项目主要研究内容包括:利用稀疏学习模型填充缺失数据,非负稀疏模型建立语义哈希,稀疏学习模型离群点检测和视频贴签等。在每一个实际应用,我们提出新的稀疏学习模型,然后设计有效优化方法去求解所得到的目标函数。
中文关键词: 稀疏学习;多媒体数据;数据挖掘;机器学习;
英文摘要: With the rapid development of Web, it has become an international need to process massive data. The solutions for effectively and efficiently handling massive data can affect our daily lives. For example, dealing with the large-scale data derived from cameras and mobile phones benefits for designing more advanced instruments to be used in our daily lives. The key opint to dealing with massive data is to accurately describe massive data and then to effectively discover their knowledge accroding to their derived representation. The project plans to study these two tasks via extending existing sparse learning models, i.e., designing novel sparse learning models to senmatically represent large-scale data, as well as to effectively and efficiently mine the hidden knowledge. More specifically, we design new sparse learning models for performing missing value imputation, buiding semantic hash, tagging videos and detection outliers. In addition, for each real application, we first design new sparse learning models, and then propose new optimization solutions to the objective functions of the resulted models.
英文关键词: sparse laerning;multimedia data;data mining;machine learning;