智能存储算法及其应用研究
文献类型：学位论文
| 作者 | 夏知拓 |
| 学位类别 | 硕士 |
| 答辩日期 | 2013 |
| 授予单位 | 中国科学院上海光学精密机械研究所 |
| 导师 | 阮昊 |
| 关键词 | 智能存储 智能算法 层次时间记忆 基于内容图像检索 智能预测 |
| 其他题名 | Studying on the intelligent storage algorithm and its application |
| 中文摘要 | 智能存储技术作为当前计算机存储领域最前沿的技术之一，除了能使存储系统具备数据存储功能外，还要使存储系统完成一定的运算和控制功能，使得存储系统本身即可进行存储管理，即智能存储使得计算机存储系统本身具备一定的智能特征。随着移动互联网的兴起，如何应对日益增长的海量数据和非结构化信息的存储已经成为当今计算机存储领域面临的重要挑战，也对智能存储领域的相关技术提出了更高的要求。 智能存储算法作为智能存储领域中的核心技术，其效率直接关系到智能存储系统的性能。智能存储算法最主要的任务是对非结构化信息进行分类和利用历史数据进行预测，为了满足现实需要，新型仿脑智能算法已经成为智能存储算法研究领域中的热点。 本论文以HTM（Hierarchical Temporal Memory，层次时间记忆）这一新型仿脑智能算法作为研究切入点，首先介绍了HTM算法的理论基础；其次详细阐述了基于HTM算法的图像理解系统的实现方案、提出了一种基于HTM算法的肌电信号识别系统，介绍了利用多HTM识别器实现HTM算法持续学习的方法，讨论了基于多HTM识别器的CBIR（Content-based image retrieval，基于内容的图像检索）系统的架构设计和各模块的具体实现，并利用Caltech101标准图像库和互联网图像库对该系统的检索性能进行了测试。最后，本论文提出了一种利用HTM算法对随机时间序列进行预测的方法，并对该方法的性能进行了测试。 |
| 英文摘要 | Intelligent Storage is one of the most cutting-edge technology in the field of computer storage, it not only make the storage system with data storage capabilities, but also to make the storage system with features of operation and control, which provide the storage system with the storage management capabilities itself, it means that Intelligent Storage technology make the computer storage system have some intelligent features. With the rise of the mobile Internet, how to cope with the growing mountain of data and unstructured information have become the enormous challenges of computer storage system nowadays and also puts forward higher requirements for the Intelligent Storage-related technologies. As the core technology in the field of Intelligent Storage, the efficiency of the intelligent storage algorithm is directly related to the performance of intelligent storage system. The main tasks of the intelligent storage algorithm are classifying the unstructured information and makes prediction with historic data. In order to meet practical needs, brain-inspired algorithm has become a hotspot for research. This paper firstly introduces the theoretical basis of the HTM(Hierarchical Temporal Memory) algorithm; secondly presents the implementation of the image-understanding system using HTM algorithm, then proposed an EMG(Electromyography) identification system based on HTM algorithm, and introduced an method to make HTM algorithm with the capacity of continuous learning using multiple HTM classifiers in detail, and presents the design ideas and framework of the CBIR system using multiple HTM classifiers and tested this CBIR system based on Caltech 101 image set and Internet image set. At last, this paper gives out a stochastic time series predicting approach based on HTM algorithm and tested this approach. |
| 语种 | 中文 |
| 源URL | [http://ir.siom.ac.cn/handle/181231/16767]  |
| 专题 | 上海光学精密机械研究所_学位论文 |
| 推荐引用方式 GB/T 7714 | 夏知拓. 智能存储算法及其应用研究[D]. 中国科学院上海光学精密机械研究所. 2013. |
入库方式： OAI收割
来源：上海光学精密机械研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。