Cross-architecture universal feature coding via distribution alignment

2025-06-15

AI Summary

This paper addresses the challenge of unifying feature representations across heterogeneous architectures (e.g., CNNs and Transformers), where architectural disparities hinder consistent encoding. To this end, we propose a novel task—Cross-Architecture Universal Feature Coding (CAUFC)—aiming to encode features from diverse models using a single, shared encoder. Our method employs a two-stage distribution alignment strategy: first, heterogenous features are uniformly mapped into a 2D token format; second, statistical distributions are aligned via truncation and normalization. This approach breaks from conventional architecture-specific encoding paradigms. Evaluated on image classification, CAUFC significantly outperforms architecture-specific baselines, achieving superior rate-distortion trade-offs. Results empirically validate the feasibility and effectiveness of universal feature compression across architectural boundaries.

Technology Category

• Machine Learning: Feature Construction/Reformulation
• Computer Vision: Representation Learning for Vision
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Feature coding has become increasingly important in scenarios where semantic representations rather than raw pixels are transmitted and stored. However, most existing methods are architecture-specific, targeting either CNNs or Transformers. This design limits their applicability in real-world scenarios where features from both architectures coexist. To address this gap, we introduce a new research problem: cross-architecture universal feature coding (CAUFC), which seeks to build a unified codec that can effectively compress features from heterogeneous architectures. To tackle this challenge, we propose a two-step distribution alignment method. First, we design the format alignment method that unifies CNN and Transformer features into a consistent 2D token format. Second, we propose the feature value alignment method that harmonizes statistical distributions via truncation and normalization. As a first attempt to study CAUFC, we evaluate our method on the image classification task. Experimental results demonstrate that our method achieves superior rate-accuracy trade-offs compared to the architecture-specific baseline. This work marks an initial step toward universal feature compression across heterogeneous model architectures.

Problem

• Develop a unified codec for CNN and Transformer features
• Align feature distributions via format and value adjustments
• Improve rate-accuracy trade-offs in cross-architecture feature coding

Innovation

• Cross-architecture universal feature coding
• Two-step distribution alignment method
• Unify CNN and Transformer features