Computer Vision

Core Machine Learning

Training BatchNorm and Only BatchNorm: On the Expressive Power of Random Features in CNNs

March 19, 2021

Abstract

A wide variety of deep learning techniques from style transfer to multitask learning rely on training affine transformations of features. Most prominent among these is the popular feature normalization technique BatchNorm, which normalizes activations and then subsequently applies a learned affine transform. In this paper, we aim to understand the role and expressive power of affine parameters used to transform features in this way. To isolate the contribution of these parameters from that of the learned features they transform, we investigate the performance achieved when training only these parameters in BatchNorm and freezing all weights at their random initializations. Doing so leads to surprisingly high performance considering the significant limitations that this style of training imposes. For example, sufficiently deep ResNets reach 82% (CIFAR-10) and 32% (ImageNet, top-5) accuracy in this configuration, far higher than when training an equivalent number of randomly chosen parameters elsewhere in the network. BatchNorm achieves this performance in part by naturally learning to disable around a third of the random features. Not only do these results highlight the expressive power of affine parameters in deep learning, but—in a broader sense—they characterize the expressive power of neural networks constructed simply by shifting and rescaling random features.

Download the Paper

AUTHORS

Written by

Jonathan Frankle

David J. Schwab

Ari S. Morcos

Publisher

ICLR 2021

Research Topics

Computer Vision

Related Publications

May 12, 2026

Human & Machine Intelligence

NeuralSet: A High-Performing Python Package for Neuro-AI

Corentin Bel, Linnea Evanson, Julien Gadonneix, Andrea Santos Revilla, Mingfang (Lucy) Zhang, Julie Bonnaire, Charlotte Caucheteux, Alexandre Défossez, Théo Desbordes, Pablo Diego-Simón, Shubh Khanna, Juliette Millet, Pierre Orhan, Saarang Panchavati, Antoine Ratouchniak, Alexis Thual, Hubert Jacob Banville, Jarod Levy, Jean Remi King, Josephine Raugel, Jérémy Rapin, Katelyn Begany, Marlene Careil, Simon Dahan, Sophia Houhamdi, Stéphane d'Ascoli, Teon Brooks, Yohann Benchetrit

May 12, 2026

February 27, 2026

Human & Machine Intelligence

Unified Vision–Language Modeling via Concept Space Alignment

Yifu Qiu, Holger Schwenk, Paul-Ambroise Duquenne

February 27, 2026

February 11, 2026

Computer Vision

UniT: Unified Multimodal Chain-of-Thought Test-time Scaling

Leon Liangyu Chen, Haoyu Ma, Ziqi Huang, Xiaoliang Dai, Jialiang Wang, Zecheng He, Jianwei Yang, Chunyuan Li, Serena Yeung-Levy, Animesh Sinha, Chu Wang, Felix Juefei-Xu, Junzhe Sun, Zhipeng Fan

February 11, 2026

December 18, 2025

Computer Vision

Pixel Seal: Adversarial-only training for invisible image and video watermarking

Alexandre Mourachko, Hady Elsahar, Pierre Fernandez, Sylvestre Rebuffi, Tom Sander, Tomáš Souček, Tuan Tran, Valeriu Lacatusu

December 18, 2025

June 11, 2019

Computer Vision

ELF OpenGo: An Analysis and Open Reimplementation of AlphaZero | Facebook AI Research

Yuandong Tian, Jerry Ma, Qucheng Gong, Shubho Sengupta, Zhuoyuan Chen, James Pinkerton, Larry Zitnick

June 11, 2019

April 30, 2018

NLP

Computer Vision

Mastering the Dungeon: Grounded Language Learning by Mechanical Turker Descent | Facebook AI Research

Zhilin Yang, Saizheng Zhang, Jack Urbanek, Will Feng, Alexander H. Miller, Arthur Szlam, Douwe Kiela, Jason Weston

April 30, 2018

October 10, 2016

Speech & Audio

Computer Vision

Polysemous Codes | Facebook AI Research

Matthijs Douze, Hervé Jégou, Florent Perronnin

October 10, 2016

June 18, 2018

Speech & Audio

Computer Vision

Low-shot learning with large-scale diffusion | Facebook AI Research

Matthijs Douze, Arthur Szlam, Bharath Hariharan, Hervé Jégou

June 18, 2018

Help Us Pioneer The Future of AI

We share our open source frameworks, tools, libraries, and models for everything from research exploration to large-scale production deployment.