Timezone: »

 
Using quadratic equations for overparametrized models
Shuang Li · William Swartworth · Martin Takac · Deanna Needell · Robert Gower

@
in Order up! The Benefits of Higher-Order Optimization in Machine Learning »

Recently the SP (Stochastic Polyak step size) method has emerged as a competitive adaptive method for setting the step sizes of SGD. SP can be interpreted as a method specialized to interpolated models, since it solves the \emph{interpolation equations}. SP solves these equation by using local linearizations of the model. We take a step further and develop a method for solving the interpolation equations that uses the local second-order approximation of the model. Our resulting method SP2 uses Hessian-vector products to speed-up the convergence of SP. Furthermore, and rather uniquely among second-order methods, the design of SP2 in no way relies on positive definite Hessian matrices or convexity of the objective function. We show SP2 is very competitive on matrix completion, non-convex test problems and logistic regression. We also provide a convergence theory on sums-of-quadratics.

Author Information

Shuang Li (University of California, Los Angeles)
William Swartworth (University of California Los Angeles)
Martin Takac (Mohamed bin Zayed University of Artificial Intelligence (MBZUAI))
Deanna Needell (UCLA)
Robert Gower (Flatiron Institute)

More from the Same Authors