On Nonconvex Optimization for Machine Learning: Gradients, Stochasticity, and Saddle Points
This addresses a critical bottleneck in scaling optimization algorithms for high-dimensional machine learning applications, though it is an incremental improvement on existing methods.
The paper tackles the problem of gradient descent and stochastic gradient descent converging to saddle points in nonconvex optimization for machine learning, showing that perturbed versions achieve polylogarithmic dimension dependence and converge to second-order stationary points as efficiently as to first-order ones.
Gradient descent (GD) and stochastic gradient descent (SGD) are the workhorses of large-scale machine learning. While classical theory focused on analyzing the performance of these methods in convex optimization problems, the most notable successes in machine learning have involved nonconvex optimization, and a gap has arisen between theory and practice. Indeed, traditional analyses of GD and SGD show that both algorithms converge to stationary points efficiently. But these analyses do not take into account the possibility of converging to saddle points. More recent theory has shown that GD and SGD can avoid saddle points, but the dependence on dimension in these analyses is polynomial. For modern machine learning, where the dimension can be in the millions, such dependence would be catastrophic. We analyze perturbed versions of GD and SGD and show that they are truly efficient---their dimension dependence is only polylogarithmic. Indeed, these algorithms converge to second-order stationary points in essentially the same time as they take to converge to classical first-order stationary points.