{"paper":{"title":"CPCANet: Deep Unfolding Common Principal Component Analysis for Domain Generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Unfolding the Flury-Gautschi algorithm into neural layers lets common principal component analysis discover domain-invariant subspaces directly inside end-to-end training.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Abd-Krim Seghouane, Yu-Hsi Chen","submitted_at":"2026-05-06T17:09:34Z","abstract_excerpt":"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"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"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.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the common principal components discovered by the unfolded Flury-Gautschi algorithm represent truly domain-invariant features that remain useful for the downstream task without losing critical information.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CPCANet deep-unfolds Common PCA to learn domain-invariant subspaces, achieving state-of-the-art zero-shot domain generalization on standard benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Unfolding the Flury-Gautschi algorithm into neural layers lets common principal component analysis discover domain-invariant subspaces directly inside end-to-end training.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"17f9fccccf46861731fc79d17de98564b5ece2d11167f40892b8df9df21678ec"},"source":{"id":"2605.05136","kind":"arxiv","version":3},"verdict":{"id":"d9d9376c-40cc-4d05-ba36-8e054d4b5963","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T17:03:23.602282Z","strongest_claim":"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.","one_line_summary":"CPCANet deep-unfolds Common PCA to learn domain-invariant subspaces, achieving state-of-the-art zero-shot domain generalization on standard benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the common principal components discovered by the unfolded Flury-Gautschi algorithm represent truly domain-invariant features that remain useful for the downstream task without losing critical information.","pith_extraction_headline":"Unfolding the Flury-Gautschi algorithm into neural layers lets common principal component analysis discover domain-invariant subspaces directly inside end-to-end training."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.05136/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T10:36:11.817337Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T21:31:19.497039Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T13:49:12.022010Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"e23b3a6db90cb96f80dbdac031e51b05477839bdfe84cdbb04c218f3f197c6ac"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}