New Bounds For Distributed Mean Estimation and Variance Reduction

Authors: Peter Davies, Vijaykrishna Gurunanthan, Niusha Moshrefi, Saleh Ashkboos, Dan Alistarh

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

Reproducibility Variable Result LLM Response
Research Type Experimental We show experimentally that our method yields practical improvements for common applications, relative to prior approaches.
Researcher Affiliation Collaboration Peter Davies IST Austria peter.davies@ist.ac.atVijaykrishna Gurunathan IIT Bombay krishnavijay1999@gmail.comNiusha Moshrefi IST Austria niusha.moshrefi@ist.ac.atSaleh Ashkboos IST Austria saleh.ashkboos@ist.ac.atDan Alistarh IST Austria & Neural Magic dan.alistarh@ist.ac.at
Pseudocode No The paper describes algorithms in prose, such as "The simplest version of our lattice quantization algorithm can be described as follows", but does not provide structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any explicit statements about releasing source code or links to a code repository.
Open Datasets No The paper does not provide concrete access information (link, DOI, repository, or formal citation with authors/year) for a publicly available or open dataset. It mentions generating synthetic data and references other datasets in the full version without providing access details.
Dataset Splits No The paper does not specify exact dataset split percentages or absolute sample counts for training, validation, or test sets.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, memory amounts, or detailed computer specifications used for running its experiments.
Software Dependencies No The paper does not list specific software dependencies with version numbers (e.g., library names like PyTorch 1.9, or specific solver versions).
Experiment Setup Yes Figure 1: Gradient quantization results for the regression example. S = 8192, n = 2 d = 100, batch_size = 4096... Figure 1 (right) Regression convergence: S = 8192, n = 2 d = 100, lr = 0.8, batch = 4096, qlevel = 8... Figure 2: Local SGD Convergence: S = 8192, n = 2 d = 100, lr = 0.1, batch = 4096, q = 8, rep = 10