STORM+: Fully Adaptive SGD with Recursive Momentum for Nonconvex Optimization

Authors: Kfir Levy, Ali Kavis, Volkan Cevher

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical In this work we propose STORM+, a new method that is completely parameter-free, does not require large batch-sizes, and obtains the optimal O(1/T 1/3) rate for finding an approximate stationary point. Our work builds on the STORM algorithm, in conjunction with a novel approach to adaptively set the learning rate and momentum parameters.
Researcher Affiliation Academia Kfir Y. Levy Technion kfirylevy@technion.ac.il Ali Kavis EPFL ali.kavis@epfl.ch Volkan Cevher EPFL volkan.cevher@epfl.ch
Pseudocode Yes We describe our method in Alg. 1 and Eq. (8), and state its guarantees in Theorem 1. For completeness we present our method in Alg. 1
Open Source Code No The paper does not mention providing open-source code for the described methodology. It focuses purely on theoretical contributions and algorithmic development without any implementation details or links.
Open Datasets No This is a theoretical paper and does not involve experiments or the use of datasets for training.
Dataset Splits No This is a theoretical paper and does not involve experiments or the use of datasets, thus no training/validation/test splits are mentioned.
Hardware Specification No The paper is theoretical and does not describe any experiments, thus no hardware specifications are provided.
Software Dependencies No The paper is theoretical and does not describe any experiments or implementations requiring specific software dependencies with version numbers.
Experiment Setup No The paper is theoretical and focuses on algorithm design and convergence analysis. It does not describe any experimental setup details such as hyperparameters or training configurations.