Convergence and Recovery Guarantees of Unsupervised Neural Networks for Inverse Problems

arXiv:2309.12128v3 Announce Type: replace
Abstract: Neural networks have become a prominent approach to solve inverse problems in recent years. While a plethora of such methods was developed to solve inverse problems empirically, we are still lacking clear theoretical guarantees for these methods. On the other hand, many works proved convergence to optimal solutions of neural networks in a more general setting using overparametrization as a way to control the Neural Tangent Kernel. In this work we investigate how to bridge …

abstract arxiv become clear convergence cs.lg networks neural networks recovery solve type unsupervised