About

Analysis of optimization landscapes in inverse problems for fixed CNNs. This project investigates how different optimization algorithms (SGD, momentum, projected gradient, etc.) recover inputs by minimizing reconstruction error through a non-convex neural network.