How Does Batch Normalization Help Optimization?

Authors: Shibani Santurkar, Dimitris Tsipras, Andrew Ilyas, Aleksander Madry

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We provide an empirical demonstration of these findings as well as their theoretical justification. To this end, we start by investigating the connection between ICS and Batch Norm. Specifically, we consider first training a standard VGG [26] architecture on CIFAR-10 [15] with and without Batch Norm. As expected, Figures 1(a) and (b) show a drastic improvement, both in terms of optimization and generalization performance, for networks trained with Batch Norm layers.
Researcher Affiliation Academia Shibani Santurkar MIT shibani@mit.edu Dimitris Tsipras MIT tsipras@mit.edu Andrew Ilyas MIT ailyas@mit.edu Aleksander M adry MIT madry@mit.edu
Pseudocode No No pseudocode or algorithm blocks are present in the paper.
Open Source Code No The paper does not provide any statement or link indicating that the source code for the methodology is openly available.
Open Datasets Yes Specifically, we consider first training a standard VGG [26] architecture on CIFAR-10 [15] with and without Batch Norm.
Dataset Splits No The paper mentions training and testing on CIFAR-10 but does not specify the exact training/validation/test splits used for their experiments in the main text or Appendix A (which only describes noise injection details).
Hardware Specification No The paper does not specify the exact hardware (e.g., GPU models, CPU types) used for running the experiments.
Software Dependencies No The paper does not specify any particular software dependencies with version numbers.
Experiment Setup Yes Standard, LR=0.1 Standard + Batch Norm, LR=0.1 Standard, LR=0.5 Standard + Batch Norm, LR=0.5 (from Figure 1, illustrating specific learning rates used in experiments).