{"paper":{"title":"Robustness Evaluation of Hybrid Quantum Neural Networks under Noise Models via System-Level Error Mitigation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"The effectiveness of error mitigation in hybrid quantum neural networks varies significantly depending on the type and intensity of noise present.","cross_cats":[],"primary_cat":"quant-ph","authors_text":"Alberto Marchisio, Jean-Michel Dricot, Jesse Roberta Mingue Njiki, Muhammad Kashif, Muhammad Shafique, Nouhaila Innan","submitted_at":"2026-04-19T16:16:50Z","abstract_excerpt":"Quantum Neural Networks (QNNs) represent a promising direction within Quantum Machine Learning (QML), yet their realization on noisy intermediate-scale quantum (NISQ) devices remains constrained by decoherence, gate imperfections, crosstalk, and readout errors. This study provides a systematic evaluation of noise effects and mitigation strategies in hybrid quantum neural networks (HQNNs). Zero-Noise Extrapolation (ZNE), Digital Dynamical Decoupling (DDD), and Layerwise Richardson Extrapolation (LRE) are integrated into end-to-end QNN training pipelines developed with PennyLane, simulated under"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The impact of noise and the effect of mitigation are strongly dependent on the noise model and its strength. The model maintains comparatively strong performance under phase-flip and phase-damping noise, while substantial degradation is observed under high depolarizing and amplitude-damping noise.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the Qiskit Aer noise models and the chosen mitigation implementations in PennyLane/Mitiq accurately capture the dominant error sources and mitigation overheads that would appear on real NISQ hardware when running the same circuits.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Simulations show hybrid quantum neural networks on Iris data degrade under depolarizing and amplitude-damping noise while phase-flip and phase-damping noise are less damaging, with ZNE, DDD, LRE, and PEC providing limited mitigation that depends on noise type and strength.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"The effectiveness of error mitigation in hybrid quantum neural networks varies significantly depending on the type and intensity of noise present.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a0a2212a845676cb1d8157a15899a4604903bf560cea3ce1f00645880d44c967"},"source":{"id":"2604.17515","kind":"arxiv","version":2},"verdict":{"id":"7cbab93f-c7b4-4c3e-8ae4-4de44aaccf0d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T05:29:55.801560Z","strongest_claim":"The impact of noise and the effect of mitigation are strongly dependent on the noise model and its strength. The model maintains comparatively strong performance under phase-flip and phase-damping noise, while substantial degradation is observed under high depolarizing and amplitude-damping noise.","one_line_summary":"Simulations show hybrid quantum neural networks on Iris data degrade under depolarizing and amplitude-damping noise while phase-flip and phase-damping noise are less damaging, with ZNE, DDD, LRE, and PEC providing limited mitigation that depends on noise type and strength.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the Qiskit Aer noise models and the chosen mitigation implementations in PennyLane/Mitiq accurately capture the dominant error sources and mitigation overheads that would appear on real NISQ hardware when running the same circuits.","pith_extraction_headline":"The effectiveness of error mitigation in hybrid quantum neural networks varies significantly depending on the type and intensity of noise present."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.17515/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"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"}