Linear Convergent Decentralized Optimization with Compression

Authors: Xiaorui Liu, Yao Li, Rongrong Wang, Jiliang Tang, Ming Yan

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

Reproducibility Variable Result LLM Response
Research Type Experimental 5 NUMERICAL EXPERIMENT We consider three machine learning problems ℓ2-regularized linear regression, logistic regression, and deep neural network. The proposed LEAD is compared with QDGD (Reisizadeh et al., 2019a), Deep Squeeze (Tang et al., 2019a), CHOCO-SGD (Koloskova et al., 2019), and two non-compressed algorithms DGD (Yuan et al., 2016) and NIDS (Li et al., 2019).
Researcher Affiliation Academia 1 Department of Computer Science and Engineering 2 Department of Mathematics 3 Department of Computational Mathematics, Science and Engineering Michigan State University, East Lansing, MI 48823, USA
Pseudocode Yes Algorithm 1 LEAD Input: Stepsize η, parameter (α, γ), X0, H1, D1 = (I W)Z for any Z Output: XK or 1/n Pn i=1 XK i
Open Source Code No The paper does not provide any explicit statement or link regarding the release of open-source code for the described methodology.
Open Datasets Yes Logistic regression. We further consider a logistic regression problem on the MNIST dataset. Neural network. We empirically study the performance of LEAD in optimizing deep neural network by training Alex Net (240 MB) on CIFAR10 dataset.
Dataset Splits No The paper mentions using MNIST and CIFAR10 datasets for training and testing but does not explicitly provide the specific percentages or sample counts for train/validation/test splits, nor does it refer to standard predefined splits with citations for reproducibility.
Hardware Specification No The paper does not provide any specific hardware details such as GPU/CPU models, processor types, or memory amounts used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies (e.g., library names with version numbers like Python 3.8, PyTorch 1.9, or specific solver versions) needed to replicate the experiment.
Experiment Setup Yes For all experiments, we tune the stepsize η from {0.01, 0.05, 0.1, 0.5}. For QDGD, CHOCO-SGD and Deepsqueeze, γ is tuned from {0.01, 0.1, 0.2, 0.4, 0.6, 0.8, 1.0}. ... For LEAD, we simply fix α = 0.5 and γ = 1.0 for all experiments... The mini-batch size is 64 for each agents.