|
|
论文:2017,Vol:35,Issue(5):786-792 |
|
|
引用本文： |
|
|
蒋长鸿, 范钢龙. 先验信息优化的S3VM算法模型研究[J]. 西北工业大学学报 |
|
|
Jiang Changhong, Fan Ganglong. A Novel Algorithm Model Design of S3VM Improved by Prior Information[J]. Northwestern polytechnical university |
|
|
|
|
|
|
|
| 先验信息优化的S3VM算法模型研究 |
|
| 蒋长鸿1, 范钢龙2,3 |
|
1. 西北工业大学 计算机学院, 陕西 西安 710072;
2. 洛阳师范学院 电子商务学院, 河南 洛阳 471934;
3. 河南省电子商务大数据处理与分析重点实验室, 河南 洛阳 471934 |
| 摘要: |
| 在没有充足的标记数据支撑的情况下，传统的半监督学习算法的运算精确性和效率会显著降低。因此，在大数据应用中，面对海量增长的数据和其中所包含的庞杂信息，需要对传统算法进行优化。但相对于数据特征，数据的先验信息是数据集中的一个未被重视的重要组成部分。文中建立了合适的数学模型引入先验信息；针对静态先验信息、动态先验信息2种不同形式先验信息，对传统S3VM算法进行了优化。测试实验验证了模型的效果。 |
| 关键词:
半监督学习
支持向量机
先验信息
|
|
| A Novel Algorithm Model Design of S3VM Improved by Prior Information |
|
| Jiang Changhong1, Fan Ganglong2,3 |
|
1. School of Computer Science and Engineering, Northwestern Polytechnical University, Xi'an 710072, China;
2. School of Electronic Commerce, Luoyang Normal University, Luoyang 471934, China;
3. Key Lab on E-Commercial Big-Data Processing and Analysis of Henan, Luoyang 471934, China |
| Abstract: |
| If there is insufficient labeled data to support, the traditional semi-supervised learning algorithm has a significant reduction in computational accuracy and efficiency. Therefore, facing the massive growth of data and the included complex information, traditional algorithms must be improved for the applications of Big-date. The prior information of data, comparing with the data feature, is an important component of the data set that hasn't been focused on. This paper establishes an appropriate mathematical model to introduce the priori information, and designs two models to improve the traditional S3VM algorithm by two different kind of prior information, static priori information and dynamic priori information. The test experiment verifies the effect of the model. |
| Key words:
big data
computational efficiency
MATLAB
supervised learning
semi-supervised learning
support vector machine
priori information
|
|
| 收稿日期: 2017-03-12
修回日期:
|
| DOI: |
| 基金项目: 国家自然科学基金（61373120）及航空科学基金（2014ZD53049）资助 |
| 通讯作者:
Email： |
| 作者简介: 蒋长鸿(1991-),西北工业大学硕士研究生,主要从事人工智能与模式识别算法研究。
|
|
|
|
|
|
|
|
参考文献: |
|
|
[1] Palaniappan R, Sundaraj K, Sundaraj S. A Comparative Study of the SVM and K-nn Machine Learning Algorithms for the Diagnosis of Respiratory Pathologies Using Pulmonary Acoustic Signals[J]. BMC Bio Information, 2014, 15(1):1-8
[2] Guillory A, Bilmes J A. Active Semi-Supervised Learning Using Submodular Functions[C]//Twenty-Seventh Conference on Uncertainty in Artificial Intelligence, 2011:274-282
[3] Gu Y, Feng K. Optimized Laplacian SVM with Distance Metric Learning for Hyperspectral Image Classification[J]. IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing, 2013, 6(3):1109-1117
[4] Günen M, Alpaydln E. Localized Algorithms for Multiple Kernel Learning[J]. Pattern Recognition, 2013, 46(3):795-807
[5] Fujino A, Ueda N, Nagata M. Adaptive Semi-Supervised Learning on Labeled and Unlabeled Data with Different Distributions[J]. Knowledge and Information Systems, 2013, 37(1):129-154 |
|
|
|
|
|
|
|
