Variance Reduction via Accelerated Dual Averaging for Finite-Sum Optimization

Authors: Chaobing Song, Yong Jiang, Yi Ma

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through experiments on real datasets, we show the good performance of VRADA over existing methods for large-scale machine learning problems.
Researcher Affiliation Academia Tsinghua-Berkeley Shenzhen Institute, Tsinghua University songcb16@mails.tsinghua.edu.cn, jiangy@sz.tsinghua.edu.cn Department of EECS, University of California, Berkeley yima@eecs.berkeley.edu
Pseudocode Yes Algorithm 1 Variance Reduction via Accelerated Dual Averaging (VRADA)
Open Source Code No No explicit statement or link providing concrete access to the source code for the methodology described in this paper was found.
Open Datasets Yes The datasets we use are a9a and covtype, downloaded from the Lib SVM website10. The dataset url is https://www.csie.ntu.edu.cn/~cjlin/libsvmtools/datasets/.
Dataset Splits No No explicit information regarding training/validation/test dataset splits (e.g., percentages, sample counts, or specific cross-validation setup) was found.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory specifications, or cloud instances) used for running the experiments were provided.
Software Dependencies No No specific software dependencies with version numbers (e.g., library names like PyTorch 1.9 or specific solver versions) were provided in the paper.
Experiment Setup Yes The problem we study is the 2-norm regularized logistic regression problem with regularization parameter λ ∈ {0, 10−8, 10−4}. All four algorithms we compare have a similar outer-inner structure, where we set all the number of iterations as m = 2n.