We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Gradientless Descent: High-Dimensional Zeroth-Order Optimization

Abstract: Zeroth-order optimization is the process of minimizing an objective $f(x)$, given oracle access to evaluations at adaptively chosen inputs $x$. In this paper, we present two simple yet powerful GradientLess Descent (GLD) algorithms that do not rely on an underlying gradient estimate and are numerically stable. We analyze our algorithm from a novel geometric perspective and present a novel analysis that shows convergence within an $\epsilon$-ball of the optimum in $O(kQ\log(n)\log(R/\epsilon))$ evaluations, for any monotone transform of a smooth and strongly convex objective with latent dimension $k < n$, where the input dimension is $n$, $R$ is the diameter of the input space and $Q$ is the condition number. Our rates are the first of its kind to be both 1) poly-logarithmically dependent on dimensionality and 2) invariant under monotone transformations. We further leverage our geometric perspective to show that our analysis is optimal. Both monotone invariance and its ability to utilize a low latent dimensionality are key to the empirical success of our algorithms, as demonstrated on BBOB and MuJoCo benchmarks.
Comments: 11 main pages, 26 total pages
Subjects: Machine Learning (cs.LG); Optimization and Control (math.OC); Machine Learning (stat.ML)
Journal reference: ICLR 2020 Spotlight
Cite as: arXiv:1911.06317 [cs.LG]
  (or arXiv:1911.06317v4 [cs.LG] for this version)

Submission history

From: Qiuyi Zhang [view email]
[v1] Thu, 14 Nov 2019 18:58:13 GMT (1317kb,D)
[v2] Fri, 15 Nov 2019 01:23:11 GMT (1317kb,D)
[v3] Tue, 19 Nov 2019 01:16:07 GMT (1317kb,D)
[v4] Mon, 18 May 2020 20:08:57 GMT (1317kb,D)

Link back to: arXiv, form interface, contact.