摘要
通过分析LCNN的学习方程,发现Lagrange约束项的物理本质是有监督学习的下降速率,提出了自适应LCNN(ALCNN)算法,避开了病态矩阵的问题,并将学习矩阵和独立成分求解复杂性都降到了O(n)。
In this paper, LCNN equation is investigated carefully and the inward nature of constraints, which was the down speed of supervised learning, is discovered. At the end, adaptive LCNN （ALCNN） is proposed, which not only can solve ill-conditioned matrix, but also the computing complexities of learning matrix and independent components are sympolied to O（n）.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2007年第1期11-13,共3页
Journal of University of Electronic Science and Technology of China
关键词
独立成分分析
盲源分离系统
鸡尾酒会问题
independent component analysis
blind source separation
cocktail party problem
