Random Scaling and Momentum for Non-smooth Non-convex Optimization

Authors: Qinzi Zhang, Ashok Cutkosky

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To validate this finding empirically, we implemented the SGDM algorithm with random scaling and assessed its performance against the standard SGDM optimizer without random scaling. Our evaluation involved the ResNet-18 model (He et al., 2016) on the CIFAR-10 image classification benchmark (Krizhevsky & Hinton, 2009).
Researcher Affiliation Academia 1Department of Electrical and Computer Engineering, Boston University, Boston, USA.
Pseudocode Yes Algorithm 1 O2NC (Cutkosky et al., 2023) and Algorithm 2 Exponentiated O2NC
Open Source Code No The paper does not contain an explicit statement about releasing source code or a link to a code repository for the methodology described.
Open Datasets Yes Our evaluation involved the ResNet-18 model (He et al., 2016) on the CIFAR-10 image classification benchmark (Krizhevsky & Hinton, 2009).
Dataset Splits No The paper mentions using the CIFAR-10 dataset and reports train and test metrics, but does not explicitly specify the data splits (e.g., percentages or counts for training, validation, and testing sets) or refer to standard predefined splits for reproducibility.
Hardware Specification No The paper does not provide specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments.
Software Dependencies No The paper does not provide specific version numbers for software dependencies or libraries used in the experiments.
Experiment Setup Yes For the hyperparameters, we configured the learning rate at 0.01, the momentum constant at 0.9, and the weight decay at 5e-4.