Gradient Descent Converges to Minimizers
arXiv:1602.04915v2215 citations
Originality Incremental advance
AI Analysis
This addresses the convergence behavior of gradient descent for optimization in machine learning, providing a theoretical guarantee for practitioners.
The paper proves that gradient descent converges to a local minimizer with random initialization, establishing this result almost surely using the Stable Manifold Theorem from dynamical systems theory.
We show that gradient descent converges to a local minimizer, almost surely with random initialization. This is proved by applying the Stable Manifold Theorem from dynamical systems theory.