On the Hyperparameters in Stochastic Gradient Descent with Momentum

Authors: Bin Shi

JMLR 2024 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental An experimental comparison between SGD and SGD with momentum is shown in Figure 1. Figure 1: The comparison for the training error between SGD and SGD with momentum. The setting is a 20-layer convolutional neural network on CIFAR-10 (Krizhevsky, 2009) with a mini-batch size of 128. Learning Rate: s = 0.01. Momentum Coefficient: α = 0.9.
Researcher Affiliation Academia Bin Shi EMAIL Academy of Mathematics and Systems Science Chinese Academy of Sciences Beijing, 100190, China School of Mathematical Science University of Chinese Academy of Sciences Beijing 100049, China
Pseudocode No The paper describes algorithms like SGD with momentum and Nesterov momentum using mathematical equations (e.g., xk+1 = xk s f(xk) + sξk + α(xk xk 1)) but does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide concrete access to source code for the methodology described. There are no explicit statements about code release, repository links, or code in supplementary materials.
Open Datasets Yes The setting is a 20-layer convolutional neural network on CIFAR-10 (Krizhevsky, 2009) with a mini-batch size of 128.
Dataset Splits No The paper mentions using 'CIFAR-10' and a 'mini-batch size of 128' but does not explicitly provide training/test/validation dataset splits or refer to predefined splits with citations for reproducibility of data partitioning.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, processor types, memory amounts) used for running its experiments. It only mentions 'a 20-layer convolutional neural network' which implies computation but no hardware specifics.
Software Dependencies No The paper discusses the theoretical analysis of optimization algorithms and mentions deep learning in the context of experiments, but it does not specify any software names with version numbers (e.g., Python, PyTorch, TensorFlow) that were used to implement the methods or run the experiments.
Experiment Setup Yes The setting is a 20-layer convolutional neural network on CIFAR-10 (Krizhevsky, 2009) with a mini-batch size of 128. Learning Rate: s = 0.01. Momentum Coefficient: α = 0.9.