We develop simple differentially private optimization algorithms that move along directions of (expected) descent to find an approximate second-order solution for nonconvex ERM. We use line search, mini-batching, and a two-phase strategy to improve the speed and practicality of the algorithm. Numerical experiments demonstrate the effectiveness of these approaches.
翻译:我们开发简单的、有差别的私人优化算法,沿着(预期的)下降方向前进,为非电流机构风险管理找到一个大致的二阶解决方案。 我们使用线搜索、微型连接和两阶段战略来提高算法的速度和实用性。 数字实验证明了这些方法的有效性。