The troublesome kernel: why deep learning for inverse problems is typically unstable.
Nina M. GottschlingVegard AntunBen AdcockAnders C. HansenPublished in: CoRR (2020)
Keyphrases
- deep learning
- inverse problems
- global optimization
- image reconstruction
- convex optimization
- unsupervised learning
- optimization methods
- optimization problems
- machine learning
- mental models
- partial differential equations
- early vision
- weakly supervised
- feature space
- super resolution
- computationally expensive
- smoothness constraint
- pairwise
- face recognition
- computer vision