Title :
Fully-multiplicative orthogonal-group ICA neural algorithm
Author_Institution :
Fac. of Eng., Perugia Univ., Terni, Italy
Abstract :
A fully-multiplicative, batch learning algorithm for neural independent component analysis (ICA) on the orthogonal group O(N) is presented. It exploits the known principle of diagonalisation of a tensor of a warped network´s outputs, which is achieved through fast iteration.
Keywords :
independent component analysis; iterative methods; neural nets; source separation; tensors; ICA; batch learning algorithm; fast iteration; fully-multiplicative orthogonal-group ICA; neural independent component analysis; statistically independent signals; warped network outputs tensor diagonalisation;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:20031108
