Global convergence of gradient descent for phase retrieval
arXiv:2410.09990v21 citationsh-index: 2
Originality Incremental advance
AI Analysis
This addresses the challenge of non-convex optimization in phase retrieval, which is incremental as it builds on existing methods to improve convergence guarantees.
The paper tackled the problem of ensuring global convergence in phase retrieval by proposing a tensor-based criterion for benign landscape and establishing boundedness of gradient trajectories, resulting in gradient descent converging to a global minimum for almost every initial point.
We propose a tensor-based criterion for benign landscape in phase retrieval and establish boundedness of gradient trajectories. This implies that gradient descent will converge to a global minimum for almost every initial point.