Toward Multi-Domain and Long-Tailed Quantization via Feature Alignment and Scaling

arXiv 2606.04920 ↗First posted 3 Jun 2026cs.LG cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Quantizing deep neural networks is essential for efficient inference on resource-constrained devices. However, most existing methods are designed for single-domain and class-balanced data, leaving practical settings with domain shifts or severe class imbalance underexplored. We address these challenges with Efficient Multi-Domain Alignment Quantization (EmaQ), which aligns domain distributions through a CDF-based projection and uses sensitivity-aware weight aggregation to stabilize multi-domain quantization. We further extend EmaQ to EmaQ-LT for long-tailed quantization by introducing class-conditioned variance scaling and confidence-based logit adjustment to mitigate majority-class overconfidence. Theoretical analyses establish convergence guarantees and motivate the proposed sensitivity and scaling mechanisms. Experiments on standard, multi-domain (Office-31, Digits), and long-tailed (SynDigits-LT, CIFAR-10-LT, CIFAR-100-LT) benchmarks show that EmaQ and EmaQ-LT achieve strong low-bit performance under domain shift and class imbalance.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 52 of its 89 references that are in the map. First posted on arXiv 3 Jun 2026.

Builds on

52 of 89 references.

References · 52

2015Deep Residual Learning for Image Recognitioncited by 52,096
2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
2016Densely Connected Convolutional Networkscited by 6,720
2018MobileNetV2: Inverted Residuals and Linear Bottleneckscited by 4,795
2015Domain-Adversarial Training of Neural Networkscited by 3,044
2019Class-Balanced Loss Based on Effective Number of Samplescited by 1,245
2019Learning Imbalanced Datasets with Label-Distribution-Aware Margin Losscited by 989
2021A Survey of Quantization Methods for Efficient Neural Network Inferencecited by 815
2019Decoupling Representation and Classifier for Long-Tailed Recognitioncited by 809
2019Large-Scale Long-Tailed Recognition in an Open Worldcited by 705
2019Learned Step Size Quantizationcited by 628

38 more

2020Up or Down? Adaptive Rounding for Post-Training Quantizationcited by 621
2016DoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low Bitwidth Gradientscited by 569
2021Deep Long-Tailed Learning: A Surveycited by 417
2020ZeroQ: A Novel Zero Shot Quantization Frameworkcited by 269
2020Disentangling Label Distribution for Long-tailed Visual Recognitioncited by 172
2019Quantization Networkscited by 153
2019Improving Neural Network Quantization without Retraining using Outlier Channel Splittingcited by 151
2019Additive Powers-of-Two Quantization: An Efficient Non-uniform Discretization for Neural Networkscited by 148
2019Transferrable Prototypical Networks for Unsupervised Domain Adaptationcited by 141
2017Deep Learning with Low Precision by Half-wave Gaussian Quantizationcited by 117
2019Dynamic Curriculum Learning for Imbalanced Data Classificationcited by 93
2020Generative Low-bitwidth Data Free Quantizationcited by 88
2020Data-Free Network Quantization With Adversarial Knowledge Distillationcited by 80
2022PD-Quant: Post-Training Quantization based on Prediction Difference Metriccited by 73
2022SQuant: On-the-Fly Data-Free Quantization via Diagonal Hessian Approximationcited by 55
2023Towards Accurate Post-training Quantization for Diffusion Modelscited by 47
2021Zero-shot Adversarial Quantizationcited by 44
2021IntraQ: Learning Synthetic Images with Intra-Class Heterogeneity for Zero-Shot Network Quantizationcited by 43
2018Post-training 4-bit quantization of convolution networks for rapid-deploymentcited by 42
2021Qimera: Data-free Quantization with Synthetic Boundary Supporting Samplescited by 40
2024LTGC: Long-tail Recognition via Leveraging LLMs-driven Generated Contentcited by 33
2024MagR: Weight Magnitude Reduction for Enhancing Post-Training Quantizationcited by 28
2024Instance-Aware Group Quantization for Vision Transformerscited by 25
2024Quantization without Tearscited by 25
2021BatchQuant: Quantized-for-all Architecture Search with Robust Quantizercited by 20
2022It's All In the Teacher: Zero-Shot Quantization Brought Closer to the Teachercited by 20
2023Rethinking the Evaluation Protocol of Domain Generalizationcited by 20
2025APHQ-ViT: Post-Training Quantization with Average Perturbation Hessian Based Reconstruction for Vision Transformerscited by 19
2022Genie: Show Me the Data for Quantizationcited by 17
2021Generative Zero-shot Network Quantizationcited by 15
2019Contrastively Smoothed Class Alignment for Unsupervised Domain Adaptationcited by 14
2025Automatic Joint Structured Pruning and Quantization for Efficient Neural Network Training and Compressioncited by 14
2021Quantization of Deep Neural Networks for Accurate Edge Computingcited by 13
2023Rethinking Data-Free Quantization as a Zero-Sum Gamecited by 13
2024QT-DoG: Quantization-aware Training for Domain Generalizationcited by 13
2023EQ-Net: Elastic Quantization Neural Networkscited by 10
2024Towards Next-Level Post-Training Quantization of Hyper-Scale Transformerscited by 9
2024MetaAug: Meta-Data Augmentation for Post-Training Quantizationcited by 5

50 most cited of 52 All 52 (JSON)

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way