Saved in:
Bibliographic Details
Main Authors: Sun, Pengfei, Jiang, Wenyu, Chee, Piew Yoong, Devos, Paul, Botteldooren, Dick
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.16476
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911325761830912
author Sun, Pengfei
Jiang, Wenyu
Chee, Piew Yoong
Devos, Paul
Botteldooren, Dick
author_facet Sun, Pengfei
Jiang, Wenyu
Chee, Piew Yoong
Devos, Paul
Botteldooren, Dick
contents Quantised neural networks (QNNs) shrink models and reduce inference energy through low-bit arithmetic, yet most still depend on a running statistics batch normalisation (BN) layer, preventing true integer-only deployment. Prior attempts remove BN by parameter folding or tailored initialisation; while helpful, they rarely recover BN's stability and accuracy and often impose bespoke constraints. We present a BN-free, fully integer QNN trained via a progressive, layer-wise distillation scheme that slots into existing low-bit pipelines. Starting from a pretrained BN-enabled teacher, we use layer-wise targets and progressive compensation to train a student that performs inference exclusively with integer arithmetic and contains no BN operations. On ImageNet with AlexNet, the BN-free model attains competitive Top-1 accuracy under aggressive quantisation. The procedure integrates directly with standard quantisation workflows, enabling end-to-end integer-only inference for resource-constrained settings such as edge and embedded devices.
format Preprint
id arxiv_https___arxiv_org_abs_2512_16476
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Batch Normalization-Free Fully Integer Quantized Neural Networks via Progressive Tandem Learning
Sun, Pengfei
Jiang, Wenyu
Chee, Piew Yoong
Devos, Paul
Botteldooren, Dick
Machine Learning
Signal Processing
Quantised neural networks (QNNs) shrink models and reduce inference energy through low-bit arithmetic, yet most still depend on a running statistics batch normalisation (BN) layer, preventing true integer-only deployment. Prior attempts remove BN by parameter folding or tailored initialisation; while helpful, they rarely recover BN's stability and accuracy and often impose bespoke constraints. We present a BN-free, fully integer QNN trained via a progressive, layer-wise distillation scheme that slots into existing low-bit pipelines. Starting from a pretrained BN-enabled teacher, we use layer-wise targets and progressive compensation to train a student that performs inference exclusively with integer arithmetic and contains no BN operations. On ImageNet with AlexNet, the BN-free model attains competitive Top-1 accuracy under aggressive quantisation. The procedure integrates directly with standard quantisation workflows, enabling end-to-end integer-only inference for resource-constrained settings such as edge and embedded devices.
title Batch Normalization-Free Fully Integer Quantized Neural Networks via Progressive Tandem Learning
topic Machine Learning
Signal Processing
url https://arxiv.org/abs/2512.16476