摘要
21世纪是数据信息时代,移动互联、社交网络、电子商务大大拓展了互联网的疆界和应用领域,由此而衍生的各类数据呈爆炸式增长,使得传统的数据分析手段已无法进行有效的数据分析。为了有效解决大规模图像数据的高效检索问题,满足大规模图像数据库的实际应用需求,提出一种基于快速极限学习机自编码(ELM-AE)的哈希二进制自编码算法。算法通过ELM-AE对数据样本进行优化,提升了图像检索的效率;通过二进制哈希实现高维图像数据向低维的二进制空间的映射和重表,提高了图像检索的精度和效率;此外,通过非线性激励函数解决了线性函数在处理非线性数据时的局限。实验结果表明,基于ELM的二进制自编码哈希算法在运行时间等方面有着良好的表现,取得了良好的检索效率和精确度。
The field of Intemet applications is so expandable because of the development of mobile Interact, social network and e-com- merce in the data information age of 21 century that the various types of data are in explosive growth, which make the traditional data a- nalysis ineffective. In order to effectively solve the problem of retrieval of image with large scale and meet the application requirements of large scale image database,a binary nonlinear hashing algorithm based on Extreme Learning Machine Auto-Encoders （ELM-AE） is pro- posed. It optimizes the data sample by ELM-AE and raises the efficiency of image retrieval. Through binary hashing to implement the mapping from high-dimensional image data to low-dimensional binary space,the retrieval accuracy and efficiency are improved. In addi- tion, nonlinear retrieval problem is solved by nonlinear activation function. The experimental results show that the proposed algorithm a- chieves good retrieval efficiency and accuracy with good performance in operation time and other aspects.
出处
《计算机技术与发展》
2017年第12期61-66,共6页
Computer Technology and Development
基金
国家自然科学基金资助项目(61572399)
陕西省青年科技新星资助项目(2013KJXX-29)
关键词
哈希学习
自编码
极限学习机
图像检索
机器学习
hashing learning
auto-encoders
extreme learning machine
image retrieval
machine learning
