Learning and Design of Principal Curves
IEEE Transactions on Pattern Analysis and Machine Intelligence
Finding Curvilinear Features in Spatial Point Patterns: Principal Curve Clustering with Noise
IEEE Transactions on Pattern Analysis and Machine Intelligence
A Unified Model for Probabilistic Principal Surfaces
IEEE Transactions on Pattern Analysis and Machine Intelligence
Introduction to Algorithms
Kernel-based nonlinear blind source separation
Neural Computation
Misep—linear and nonlinear ICA based on mutual information
The Journal of Machine Learning Research
Principal curves with bounded turn
IEEE Transactions on Information Theory
Hi-index | 0.00 |
Nonlinear independent components analysis (NICA) is known to be an ill-posed problem when only the independence of the sources are sought. Additional constraints on the distribution of the sources or the structure of the mixing nonlinearity are imposed to achieve a solution that is unique in a suitable sense. In this paper, we present a technique that tackles nonlinear blind source separation (NBSS) as a nonlinear invertible coordinate unfolding problem utilizing a recently developed definition of maximum-likelihood principal curves. The proposition would be applicable most conveniently to independent unimodal source distributions with mixtures that have diminishing second order derivatives along the source axes. Application to multimodal sources would be possible with some modifications that are not discussed in this paper. The ill-posed nature of NBSS is also discussed from a differential geometric perspective in this context.