Learning Explicitly Conditioned Sparsifying Transforms
Abstract
Sparsifying transforms are widely recognized tools for discovering structured sparse representations of signals in specific transform domains. Typically, the conditioning number and representation ability are conflicting key features of learning square transforms that may not be explicitly controlled in a given optimization model. Unlike existing approaches from the literature, our paper presents a new sparsifying transform model that explicitly controls both the data representation quality and the condition number of the learned transforms. We confirm by numerical experiments that our model has better numerical behavior than the state-of-the-art methods developed in the same line of work.
Fields of Study & Subjects
Neural Networks and Applications
Publication details
- Journal / source
- ICASSP 2026 - 2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
- Publisher
- IEEE
- Pages
- 576–580
- Publication date
- 21 April 2026
- Language
- EN