-
摘要: 信息准则在发展主子空间跟踪算法中具有十分重要的意义.目前, 主子空间信息准则还不多见, 为此本文提出了一种新型的主子空间跟踪信息准则.当且仅当神经网络权矩阵收敛到信号主子空间的一个正交基时, 该信息准则取得唯一全局极大值.通过梯度上升法对该信息准则求极值导出了一个主子空间跟踪算法.最后通过仿真实验和实际应用证明了所提算法的正确性和实用性.Abstract: It is clear that a properly chosen information criterion is a very important part in developing principal subspace tracking algorithms. Up to now, the numbers of principal subspace tracking information criterion are not rich. Based on this fact, a novel information criterion is proposed in this paper. This information criterion exhibits a single global maximum attained if and only if the weight matrix spans the desired principal subspace. A subspace tracking algorithm is derived by using gradient ascent method to this information criterion. Numerical simulation and real application experiments are carried out to show the correctness and practicality of the proposed algorithm.
-
Key words:
- Principal subspace tracking /
- information criterion /
- gradient ascent method /
- neural networks
1) 本文责任编委 梅生伟 -
[1] Helwani K. Fundamentals of adaptive filter theory. Adaptive Identification of Acoustic Multichannel Systems Using Sparse Representations. Cham: Springer International Publishing, 2015. [2] Haddadi F. Steady-state statistical performance analysis of subspace tracking methods. IEEE Transactions on Signal Processing, 2016, 64(18): 4781-4791 doi: 10.1109/TSP.2016.2572039 [3] 孔令智, 高迎彬, 李红增, 张华鹏.一种快速的多个主成分并行提取算法.自动化学报, 2017, 43(5): 835-842 doi: 10.16383/j.aas.2017.c160299Kong Ling-Zhi, Gao Ying-Bin, Li Hong-Zeng, Zhang Hua-Peng. A fast algorithm that extracts multiple principle components in parallel. Acta Automatica Sinica, 2017, 43(5): 835-842 doi: 10.16383/j.aas.2017.c160299 [4] Nguyen V D, Abed-Meraim K, Linh-Trung N, Weber R. Generalized minimum noise subspace for array processing. IEEE Transactions on Signal Processing, 2017, 65(14): 3789-3802 doi: 10.1109/TSP.2017.2695457 [5] Du K L, Swamy M N S. Principal component analysis. Neural Networks and Statistical Learning. London: Springer, 2014. doi: 10.1007/978-1-4471-5571-3_12 [6] Doukopoulos X G, Moustakides G V. Fast and stable subspace tracking. IEEE Transactions on Signal Processing, 2008, 56(4): 1452-1465 doi: 10.1109/TSP.2007.909335 [7] Golub G H, Van Loan C F. Matrix computations. Mathematical Gazette, 1996, 47(5 Series Ⅱ): 392-396 http://dl.acm.org/citation.cfm?id=248979&dl=ACM&coll=portal&... [8] Wang R, Yao M L, Zhang D M, Zou H X. A novel orthonormalization matrix based fast and stable DPM algorithm for principal and minor subspace tracking. IEEE Transactions on Signal Processing, 2012, 60(1): 466-472 doi: 10.1109/TSP.2011.2169406 [9] Gao Y B, Kong X Y, Zhang H H, Hou L A. A weighted information criterion for multiple minor components and its adaptive extraction algorithms. Neural Networks, 2017, 89: 1-10 doi: 10.1016/j.neunet.2017.02.006 [10] Yang B. Projection approximation subspace tracking. IEEE Transactions on Signal Processing, 1995, 43(1): 95-107 doi: 10.1109/78.365290 [11] Miao Y, Hua Y. Fast subspace tracking and neural network learning by a novel information criterion. IEEE Transactions on Signal Processing, 1998, 46(7): 1967-1979 doi: 10.1109/78.700968 [12] Kong X Y, Hu C H, Han C Z. A dual purpose principal and minor subspace gradient flow. IEEE Transactions on Signal Processing, 2012, 60(1): 197-210 doi: 10.1109/TSP.2011.2169060 [13] Kong X Y, Hu C H, Duan Z S. Principal Component Analysis Networks and Algorithms. Singapore: Springer, 2017. -
下载:
图(6)
计量
- 文章访问数: 1173
- HTML全文浏览量: 269
- PDF下载量: 121
- 被引次数: 0
