Abstract
In this paper generalizations of Sanger's learning rule for nondefinite matrices are explored. It is shown that the left and right principal components of any matrix can be computed so that these components upper triangulize the original matrix. We also modified the original Sanger's system to obtain new dynamical systems with a larger domain of attraction. Stability analysis for several Sanger's type systems for the standard and generalized principal, and minor component analyzers applied to nonsymmetric matrices is developed.
Original language | English (US) |
---|---|
Title of host publication | 2006 IEEE Sensor Array and Multichannel Signal Processing Workshop Proceedings, SAM 2006 |
Pages | 425-429 |
Number of pages | 5 |
State | Published - 2006 |
Event | 4th IEEE Sensor Array and Multichannel Signal Processing Workshop Proceedings, SAM 2006 - Waltham, MA, United States Duration: Jul 12 2006 → Jul 14 2006 |
Publication series
Name | 2006 IEEE Sensor Array and Multichannel Signal Processing Workshop Proceedings, SAM 2006 |
---|
Other
Other | 4th IEEE Sensor Array and Multichannel Signal Processing Workshop Proceedings, SAM 2006 |
---|---|
Country/Territory | United States |
City | Waltham, MA |
Period | 7/12/06 → 7/14/06 |
Bibliographical note
Funding Information:We gratefully acknowledge the help and support of AgentLink III, in particular its Technical Fora, which not only motivated the authors to work together in producing this joint survey, but also provided the conditions for much of the discussion that we used in this paper and indeed that will guide future work in this area of research. We also acknowledge the valuable comments and suggestions provided by the anonymous referees.
Keywords
- Dynamical systems
- Generalized eigenvalue problem
- Global stability
- Minor component analysis
- Oja's learning rule
- Principal component analysis
- Sanger's learning rule