{"paper":{"title":"Resource-Element Energy Difference for Noncoherent Over-the-Air Federated Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"REED recovers signed model updates from energy differences on two orthogonal resources without needing instantaneous channel state information.","cross_cats":["cs.AI","cs.DC","cs.LG","stat.ML"],"primary_cat":"eess.SP","authors_text":"Hao Chen, Zavareh Bozorgasl","submitted_at":"2026-05-08T05:29:14Z","abstract_excerpt":"Over-the-air federated learning (OTA-FL) reduces uplink latency by aggregating client updates directly over the wireless multiple-access channel. Coherent analog aggregation realizes this idea by aligning the phases and amplitudes of simultaneously transmitted waveforms, which typically requires synchronization, instantaneous channel-state information (CSI), phase compensation, and power control. Noncoherent energy detection removes the need for phase-coherent combining, but a single energy measurement is nonnegative and, therefore, cannot represent signed model updates.\n  This paper introduce"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"REED is unbiased for the desired signed sum and admits an exact closed-form variance under Rayleigh fading. We incorporate REED into full-participation FedAvg and prove a smooth nonconvex stationarity bound... yielding the canonical (1/sqrt(T)) stationarity rate.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The analysis and experiments rely on Rayleigh fading, slow-timescale average channel power calibration, an average per-client energy budget, and either IID or moderate heterogeneity on MNIST/Fashion-MNIST; the stationarity bound assumes the aggregation gain can be scheduled to make the REED perturbation scale quadratically with the local stepsize.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"REED enables unbiased noncoherent signed aggregation in OTA-FL via energy differences on orthogonal resources, achieving closed-form variance under Rayleigh fading and standard (1/sqrt(T)) convergence in FedAvg.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"REED recovers signed model updates from energy differences on two orthogonal resources without needing instantaneous channel state information.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"86a685d6dd1ea819933b91a20b9881f2053edaeaaf31abecffa1f322dba57e49"},"source":{"id":"2605.07263","kind":"arxiv","version":2},"verdict":{"id":"545b8170-21a0-478a-9991-b3c584316d8e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-11T01:41:27.994362Z","strongest_claim":"REED is unbiased for the desired signed sum and admits an exact closed-form variance under Rayleigh fading. We incorporate REED into full-participation FedAvg and prove a smooth nonconvex stationarity bound... yielding the canonical (1/sqrt(T)) stationarity rate.","one_line_summary":"REED enables unbiased noncoherent signed aggregation in OTA-FL via energy differences on orthogonal resources, achieving closed-form variance under Rayleigh fading and standard (1/sqrt(T)) convergence in FedAvg.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The analysis and experiments rely on Rayleigh fading, slow-timescale average channel power calibration, an average per-client energy budget, and either IID or moderate heterogeneity on MNIST/Fashion-MNIST; the stationarity bound assumes the aggregation gain can be scheduled to make the REED perturbation scale quadratically with the local stepsize.","pith_extraction_headline":"REED recovers signed model updates from energy differences on two orthogonal resources without needing instantaneous channel state information."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.07263/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T17:01:19.190653Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T11:59:29.546020Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"3fbeff05d01a714656378e3aaa0a40e3b7a8e0a251fee349f61d600b11e00e1b"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"dee71191b72ca9887699d93fdebe23adb4f22213d5a3fdff0491c49ea0530819"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}