Toward Multi-Domain and Long-Tailed Quantization via Feature Alignment and Scaling
This work tackles the underexplored problem of quantizing neural networks under domain shift and class imbalance, which is critical for deploying models on resource-constrained devices in realistic settings.
EmaQ and EmaQ-LT address multi-domain and long-tailed quantization by aligning domain distributions via CDF-based projection and mitigating class imbalance via variance scaling and logit adjustment, achieving strong low-bit performance on benchmarks like Office-31, Digits, SynDigits-LT, CIFAR-10-LT, and CIFAR-100-LT.
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.