<ul><li>Researchers propose a new method for finding second-order stationary points in differentially private stochastic non-convex optimization.</li><li>Existing methods have limitations in convergence error rate and reliance on auxiliary private model selection procedures.</li><li>The proposed perturbed stochastic gradient descent framework addresses these issues by using Gaussian noise injection and gradient oracles.</li><li>Numerical experiments on real-world datasets validate the effectiveness of the new approach.</li></ul>

Second-Order Convergence in Private Stochastic Non-Convex Optimization

Discover more