{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5DDFVD7CJZG77OMBNZOVD5QPYI","short_pith_number":"pith:5DDFVD7C","schema_version":"1.0","canonical_sha256":"e8c65a8fe24e4dffb9816e5d51f60fc221fa5c532dc24449ef890d5792037a11","source":{"kind":"arxiv","id":"2507.06525","version":1},"attestation_state":"computed","paper":{"title":"AdaDPIGU: Differentially Private SGD with Adaptive Clipping and Importance-Based Gradient Updates for Deep Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["math.ST","stat.ML","stat.TH"],"primary_cat":"cs.LG","authors_text":"Fang Xie, Huiqi Zhang","submitted_at":"2025-07-09T03:53:03Z","abstract_excerpt":"Differential privacy has been proven effective for stochastic gradient descent; however, existing methods often suffer from performance degradation in high-dimensional settings, as the scale of injected noise increases with dimensionality. To tackle this challenge, we propose AdaDPIGU--a new differentially private SGD framework with importance-based gradient updates tailored for deep neural networks. In the pretraining stage, we apply a differentially private Gaussian mechanism to estimate the importance of each parameter while preserving privacy. During the gradient update phase, we prune low"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2507.06525","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-07-09T03:53:03Z","cross_cats_sorted":["math.ST","stat.ML","stat.TH"],"title_canon_sha256":"df08c8ac8e2e85ef23f6eead1403093e223d3d6d7b528ccef228b5bef9fac976","abstract_canon_sha256":"f77ec9d0e5db09f0fb087c985b50f6a884d14561477612df7c8516174678e612"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:34:17.691679Z","signature_b64":"KnEg0wXhqlTWOZgpnasCHYLqNzlxQOOp3DIcSSWjC6imNtW7uhOHbbWxbq6tqmpMhYMMJa/WM+ODve1FR1tHCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e8c65a8fe24e4dffb9816e5d51f60fc221fa5c532dc24449ef890d5792037a11","last_reissued_at":"2026-07-05T11:34:17.691132Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:34:17.691132Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"AdaDPIGU: Differentially Private SGD with Adaptive Clipping and Importance-Based Gradient Updates for Deep Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["math.ST","stat.ML","stat.TH"],"primary_cat":"cs.LG","authors_text":"Fang Xie, Huiqi Zhang","submitted_at":"2025-07-09T03:53:03Z","abstract_excerpt":"Differential privacy has been proven effective for stochastic gradient descent; however, existing methods often suffer from performance degradation in high-dimensional settings, as the scale of injected noise increases with dimensionality. To tackle this challenge, we propose AdaDPIGU--a new differentially private SGD framework with importance-based gradient updates tailored for deep neural networks. In the pretraining stage, we apply a differentially private Gaussian mechanism to estimate the importance of each parameter while preserving privacy. During the gradient update phase, we prune low"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.06525","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2507.06525/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2507.06525","created_at":"2026-07-05T11:34:17.691203+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.06525v1","created_at":"2026-07-05T11:34:17.691203+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.06525","created_at":"2026-07-05T11:34:17.691203+00:00"},{"alias_kind":"pith_short_12","alias_value":"5DDFVD7CJZG7","created_at":"2026-07-05T11:34:17.691203+00:00"},{"alias_kind":"pith_short_16","alias_value":"5DDFVD7CJZG77OMB","created_at":"2026-07-05T11:34:17.691203+00:00"},{"alias_kind":"pith_short_8","alias_value":"5DDFVD7C","created_at":"2026-07-05T11:34:17.691203+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5DDFVD7CJZG77OMBNZOVD5QPYI","json":"https://pith.science/pith/5DDFVD7CJZG77OMBNZOVD5QPYI.json","graph_json":"https://pith.science/api/pith-number/5DDFVD7CJZG77OMBNZOVD5QPYI/graph.json","events_json":"https://pith.science/api/pith-number/5DDFVD7CJZG77OMBNZOVD5QPYI/events.json","paper":"https://pith.science/paper/5DDFVD7C"},"agent_actions":{"view_html":"https://pith.science/pith/5DDFVD7CJZG77OMBNZOVD5QPYI","download_json":"https://pith.science/pith/5DDFVD7CJZG77OMBNZOVD5QPYI.json","view_paper":"https://pith.science/paper/5DDFVD7C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.06525&json=true","fetch_graph":"https://pith.science/api/pith-number/5DDFVD7CJZG77OMBNZOVD5QPYI/graph.json","fetch_events":"https://pith.science/api/pith-number/5DDFVD7CJZG77OMBNZOVD5QPYI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5DDFVD7CJZG77OMBNZOVD5QPYI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5DDFVD7CJZG77OMBNZOVD5QPYI/action/storage_attestation","attest_author":"https://pith.science/pith/5DDFVD7CJZG77OMBNZOVD5QPYI/action/author_attestation","sign_citation":"https://pith.science/pith/5DDFVD7CJZG77OMBNZOVD5QPYI/action/citation_signature","submit_replication":"https://pith.science/pith/5DDFVD7CJZG77OMBNZOVD5QPYI/action/replication_record"}},"created_at":"2026-07-05T11:34:17.691203+00:00","updated_at":"2026-07-05T11:34:17.691203+00:00"}