Estimates on the generalization error of Physics Informed Neural Networks (PINNs) for approximating PDEs II: A class of inverse problems.
Siddhartha MishraRoberto MolinaroPublished in: CoRR (2020)
Keyphrases
- generalization error
- inverse problems
- neural network
- partial differential equations
- cross validation
- model selection
- active learning
- unbiased estimator
- low error
- upper bound
- image reconstruction
- learning algorithm
- convex optimization
- training data
- global optimization
- training set
- sample size
- image processing
- level set
- optimization problems
- supervised learning
- image denoising
- high order
- optimization methods
- training error
- image enhancement
- simulated annealing
- lower bound
- objective function
- smoothness constraint
- multiscale
- machine learning