Compressive sensing with un-trained neural networks: Gradient descent finds a smooth approximation.
Reinhard HeckelMahdi SoltanolkotabiPublished in: ICML (2020)
Keyphrases
- compressive sensing
- trained neural networks
- neural network
- random projections
- neural learning
- sparse representation
- compressive sampling
- trained neural network
- rule extraction
- image representation
- artificial neural networks
- signal processing
- orthogonal matching pursuit
- linguistic rules
- black box
- feature vectors
- compressed sensing
- multiscale
- knowledge base