Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2503.11159 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917957101158400 |
|---|---|
| author | Pang, Junbiao Cai, Tianyang |
| author_facet | Pang, Junbiao Cai, Tianyang |
| contents | Quantization-Aware Training (QAT) is one of the prevailing neural network compression solutions. However, its stability has been challenged for yielding deteriorating performances as the quantization error is inevitable. We find that the sharp landscape of loss, which leads to a dramatic performance drop, is an essential factor that causes instability. Theoretically, we have discovered that the perturbations in the feature would bring a flat local minima. However, simply adding perturbations into either weight or feature empirically deteriorates the performance of the Full Precision (FP) model. In this paper, we propose Feature-Perturbed Quantization (FPQ) to stochastically perturb the feature and employ the feature distillation method to the quantized model. Our method generalizes well to different network architectures and various QAT methods. Furthermore, we mathematically show that FPQ implicitly regularizes the Hessian norm, which calibrates the smoothness of a loss landscape. Extensive experiments demonstrate that our approach significantly outperforms the current State-Of-The-Art (SOTA) QAT methods and even the FP counterparts. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_11159 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Stabilizing Quantization-Aware Training by Implicit-Regularization on Hessian Matrix Pang, Junbiao Cai, Tianyang Computer Vision and Pattern Recognition Quantization-Aware Training (QAT) is one of the prevailing neural network compression solutions. However, its stability has been challenged for yielding deteriorating performances as the quantization error is inevitable. We find that the sharp landscape of loss, which leads to a dramatic performance drop, is an essential factor that causes instability. Theoretically, we have discovered that the perturbations in the feature would bring a flat local minima. However, simply adding perturbations into either weight or feature empirically deteriorates the performance of the Full Precision (FP) model. In this paper, we propose Feature-Perturbed Quantization (FPQ) to stochastically perturb the feature and employ the feature distillation method to the quantized model. Our method generalizes well to different network architectures and various QAT methods. Furthermore, we mathematically show that FPQ implicitly regularizes the Hessian norm, which calibrates the smoothness of a loss landscape. Extensive experiments demonstrate that our approach significantly outperforms the current State-Of-The-Art (SOTA) QAT methods and even the FP counterparts. |
| title | Stabilizing Quantization-Aware Training by Implicit-Regularization on Hessian Matrix |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2503.11159 |