Dynamical behavior of Oja PCA model for non-symmetric covariance matrix

Oja´s principal component analysis (PCA) model is a well-known and powerful technique in the field of signal processing and data analysis. Dynamical behavior of Oja PCA model is an essential issue for practical applications. Existing convergence results are mainly concerned with the case of symmetric covariance matrix. How will Oja model behave when this symmetric condition is violated? In this paper, dynamic behavior of Oja model for non-symmetric covariance matrix is briefly analyzed. Asymptotical stability of trivial solution is established with the help of eigen-decomposition theorem. Most importantly, sufficient condition for the system to avoid having finite escape time is established. Simulation results are further used to illustrate the theoretical results.