一種新的特征結(jié)構(gòu)提取方法及其神經(jīng)網(wǎng)絡(luò)實現(xiàn)
A NEW METHOD FOR EIGENSTRUCTURE EXTRACTION AND ITS NEURAL NETWORK IMPLEMENTATION
-
摘要: 本文首先建立了特征結(jié)構(gòu)問題的代價函數(shù)表示,通過對代價函數(shù)求極小可以求得原始數(shù)據(jù)協(xié)方差矩陣的最大特征向量。為了求得其他特征向量,特構(gòu)造了一個協(xié)方差矩陣序列。為實現(xiàn)對代價函數(shù)求極小,可把高階神經(jīng)網(wǎng)絡(luò)引入特征結(jié)構(gòu)提取中。這種方法比較直觀,它將網(wǎng)絡(luò)穩(wěn)定時的輸出與所求協(xié)方差矩陣的主特征向量的各個分量相對應(yīng)。理論分析和計算機仿真均驗證了這種方法的正確性。Abstract: The cost function for eigenstuctures extration is discussed in detail, one can obtain the largest eigenvector by minimizing the cost function. In order to obtain the other eigenvectors, a covariance matrix series is constructed. If one compares the cost function with the energy function of a neural network, the neural network can be introduced to extract the eigenvectors. Theoretical analysis and simulations show that the proposed method is reasonable and feasible.
-
Yang Bin. Subspace Tracking Based on the Projection Approach and the Recursive Least Squares Method. Minnesota: Proc. IV of ICASSP93, 1993, 145-148.[2]Mathew G, et al.Gauss-Newton Adaptive Subspace Estimation. California: Proc. IV of ICASSP92,[3]92, 205-208.[4]Luo FarLong, Li Yan-Da. Real-Time Neural Computation of the Noise Subspace for the MUSIC[5]Algorithm. Minnesota: Proc. I of ICASSP93, 1993, 485-487.[6]Kung San Yuan.[J].et al. A Neural Network Learning Algorithm for Adaptive Principal Component Extraction(APEX). New Mexico: Proc. of ICASSP9.1990,:-. A Neural Network Learning Algorithm for Adaptive Principal Component Extraction(APEX). New Mexico: Proc. of ICASSP' target='_blank'>[7]虞水俊,等.國防科技大學(xué)學(xué)報,1995, 17(1): 16-21.[8]虞水俊,孔鐵生,梁甸農(nóng).電子學(xué)報,1995, 23(8): 109-111. -
計量
- 文章訪問數(shù): 2292
- HTML全文瀏覽量: 181
- PDF下載量: 568
- 被引次數(shù): 0
下載:
