CPCANet: Deep Unfolding Common Principal Component Analysis for Domain Generalization

2026-05-06

AI Summary

This work addresses the challenge of explicitly constructing a shared invariant subspace robust to out-of-distribution shifts in domain generalization. It proposes the first integration of Common PCA with deep unfolding by differentiably unrolling the Flury–Gautschi iterative algorithm into an end-to-end trainable neural network layer, enabling the learning of structured, cross-domain invariant subspaces. Grounded in second-order statistical modeling, the method requires no target-domain data for hyperparameter tuning and is independent of specific network architectures, offering both interpretability and broad applicability. Evaluated under a zero-shot transfer setting across four standard domain generalization benchmarks, the approach achieves state-of-the-art performance while maintaining computational efficiency.

Abstract

Domain Generalization (DG) aims to learn representations that remain robust under out-of-distribution (OOD) shifts and generalize effectively to unseen target domains. While recent invariant learning strategies and architectural advances have achieved strong performance, explicitly discovering a structured domain-invariant subspace through second-order statistics remains underexplored. In this work, we propose CPCANet, a novel framework grounded in Common Principal Component Analysis (CPCA), which unrolls the iterative Flury-Gautschi (FG) algorithm into fully differentiable neural layers. This approach integrates the statistical properties of CPCA into an end-to-end trainable framework, enforcing the discovery of a shared subspace across diverse domains while preserving interpretability. Experiments on four standard DG benchmarks demonstrate that CPCANet achieves state-of-the-art (SOTA) performance in zero-shot transfer. Moreover, CPCANet is architecture-agnostic and requires no dataset-specific tuning, providing a simple and efficient approach to learning robust representations under distribution shift. Code is available at https://github.com/wish44165/CPCANet.

Problem

• Domain Generalization
• Common Principal Component Analysis
• Out-of-Distribution Generalization
• Domain-Invariant Subspace
• Zero-Shot Transfer

Innovation

• Deep Unfolding
• Common Principal Component Analysis
• Domain Generalization
• Differentiable Algorithm
• Invariant Subspace