English

Stratified Knowledge-Density Super-Network for Scalable Vision Transformers

Machine Learning 2025-11-18 v1 Artificial Intelligence Computer Vision and Pattern Recognition

Abstract

Training and deploying multiple vision transformer (ViT) models for different resource constraints is costly and inefficient. To address this, we propose transforming a pre-trained ViT into a stratified knowledge-density super-network, where knowledge is hierarchically organized across weights. This enables flexible extraction of sub-networks that retain maximal knowledge for varying model sizes. We introduce \textbf{W}eighted \textbf{P}CA for \textbf{A}ttention \textbf{C}ontraction (WPAC), which concentrates knowledge into a compact set of critical weights. WPAC applies token-wise weighted principal component analysis to intermediate features and injects the resulting transformation and inverse matrices into adjacent layers, preserving the original network function while enhancing knowledge compactness. To further promote stratified knowledge organization, we propose \textbf{P}rogressive \textbf{I}mportance-\textbf{A}ware \textbf{D}ropout (PIAD). PIAD progressively evaluates the importance of weight groups, updates an importance-aware dropout list, and trains the super-network under this dropout regime to promote knowledge stratification. Experiments demonstrate that WPAC outperforms existing pruning criteria in knowledge concentration, and the combination with PIAD offers a strong alternative to state-of-the-art model compression and model expansion methods.

Keywords

Cite

@article{arxiv.2511.11683,
  title  = {Stratified Knowledge-Density Super-Network for Scalable Vision Transformers},
  author = {Longhua Li and Lei Qi and Xin Geng},
  journal= {arXiv preprint arXiv:2511.11683},
  year   = {2025}
}

Comments

Accepted by AAAI 2026

R2 v1 2026-07-01T07:38:07.673Z