Skip to yearly menu bar Skip to main content


Acceleration with a Ball Optimization Oracle

Yair Carmon · Arun Jambulapati · Qijia Jiang · Yujia Jin · Yin Tat Lee · Aaron Sidford · Kevin Tian

Poster Session 3 #1057


Consider an oracle which takes a point x and returns the minimizer of a convex function f in an l2 ball of radius r around x. It is straightforward to show that roughly r^{-1}\log(1/epsilon) calls to the oracle suffice to find an \epsilon-approximate minimizer of f in an l2 unit ball. Perhaps surprisingly, this is not optimal: we design an accelerated algorithm which attains an epsilon-approximate minimizer with roughly r^{-2/3} \log(1/epsilon) oracle queries, and give a matching lower bound. Further, we implement ball optimization oracles for functions with a locally stable Hessian using a variant of Newton's method and, in certain cases, stochastic first-order methods. The resulting algorithms apply to a number of problems of practical and theoretical import, improving upon previous results for logistic and linfinity regression and achieving guarantees comparable to the
state-of-the-art for l
p regression.

Chat is not available.