투영 스펙트럼 분석의 원리와 응용

초록

The objective of this paper is to define “Projection Spectral Analysis (PSA)” which generalizes Principal Component Analysis (PCA) and Independent Component Analysis (ICA), a class of unsupervised learning paradigms in artificial neural networks; to describe the mathematical backgrounds and theoretical principles; and to provide a foundation of applications by developing the associated theorems and algorithms. Here, the instances of fundamental formulas are only involved in square correlation matrices or square covariance matrices. Therefore, projection operators and nilponents, derived from the projection theorem and the reduction theorem of projection spectral analysis can be applied to numericallyt asble pattern recognition of neural networks

키워드