Natasha: Faster Non-Convex Stochastic Optimization via Strongly Non-Convex Parameter

Authors: Zeyuan Allen-Zhu

ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical Given a non-convex function f(x) that is an average of n smooth functions, we design stochastic first-order methods to find its approximate stationary points. The performance of our new methods depend on the smallest (negative) eigenvalue σ of the Hessian. This parameter σ captures how strongly non-convex f(x) is, and is analogous to the strong convexity parameter for convex optimization. At least in theory, our methods outperform known results for a range of parameter σ, and can also be used to find approximate local minima. Our result implies an interesting dichotomy: there exists a threshold σ0 so that the (currently) fastest methods for σ > σ0 and for σ < σ0 have different behaviors: the former scales with n2/3 and the latter scales with n3/4.
Researcher Affiliation Collaboration Microsoft Research. Correspondence to: Zeyuan Allen-Zhu <zeyuan@csail.mit.edu>.
Pseudocode Yes Algorithm 1 Natasha(x , p, T , α)
Open Source Code No The paper does not provide concrete access to source code (specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described in this paper.
Open Datasets No This is a theoretical paper and does not involve empirical studies with specific datasets or their public availability.
Dataset Splits No This is a theoretical paper and does not involve empirical studies with specific dataset splits.
Hardware Specification No This is a theoretical paper and does not mention specific hardware used for running experiments.
Software Dependencies No This is a theoretical paper and does not mention specific software dependencies with version numbers.
Experiment Setup No This is a theoretical paper and does not describe an experimental setup with specific hyperparameters or training configurations.