Regularization Parameter-Free Convolutional Sparse Coding via Projections Onto The ℓ1-Ball and The Discrepancy Principle.
Paul RodríguezPublished in: MLSP (2018)
Keyphrases
- sparse coding
- regularization parameter
- image restoration
- natural images
- cross validation
- sparse representation
- image classification
- linear combination
- unsupervised learning
- image representation
- generative model
- image patches
- regularization methods
- hyperparameters
- visual words
- deep belief networks
- high dimensional
- feature space
- object recognition
- probabilistic model
- model selection
- dimensionality reduction
- input image
- training set
- basis vectors
- bayesian networks
- machine learning