Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

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

About

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.

Ting-An Chen, Chin-Yuan Yeh, De-Nian Yang• 2026

Related benchmarks

TaskDatasetResultRank
Image ClassificationOffice-31
Average Accuracy80.6
357
Image ClassificationCIFAR-100 LT
Top-1 Acc55.62
225
Image ClassificationSVHN
Top-1 Accuracy96.3
209
Image ClassificationCIFAR-10-LT gamma=50 (test)
Accuracy74.4
68
Domain AdaptationMNIST to MNIST-M (test)
Acc (Target)96.1
47
Image ClassificationDigits
Accuracy (MNIST-M Source)98.78
41
Domain AdaptationMNIST to SVHN (test)
Accuracy59.9
34
Image ClassificationCIFAR-10-LT gamma=10 (test)
Accuracy81.36
34
Image ClassificationSynDigits-LT gamma=10
Accuracy96.45
32
Image ClassificationSynDigits-LT gamma=50
Accuracy89.8
31
Showing 10 of 17 rows

Other info

Follow for update