Principal Component Analysis Networks and Algorithms by Xiangyu Kong Changhua Hu & Zhansheng Duan

Author:Xiangyu Kong, Changhua Hu & Zhansheng Duan
Language: eng
Format: epub
Publisher: Springer Singapore, Singapore

(5.55)

which is the average version of the continuous-time differential equation

(5.56)

which, after discretization, gives a nonlinear stochastic learning rule

(5.57)

where denotes the learning step size, and if “+” is used, then (5.57) is a PSA algorithm, and if “−” is used, then (5.57) is a MSA algorithm. (5.25) and (5.57) constitute our unified dual-purpose principal and minor subspace gradient flow. The gradient flow (5.57) has a computational complexity of flops per update, which is cheaper than for algorithm in [22], for algorithm in [24], and for algorithm in [25]. The operations involved in (5.57) are simple matrix addition, multiplication, and inversion, which are easy for systolic array implementation [22].

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.