{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:POE6R3BMXIQXFVPNMP6UEPWKZH","short_pith_number":"pith:POE6R3BM","schema_version":"1.0","canonical_sha256":"7b89e8ec2cba2172d5ed63fd423ecac9e5ccecbd5d852ce8cdbe77d10d74d5dc","source":{"kind":"arxiv","id":"2210.00038","version":2},"attestation_state":"computed","paper":{"title":"Differentially Private Optimization on Large Model at Small Cost","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CR","cs.CV"],"primary_cat":"cs.LG","authors_text":"George Karypis, Sheng Zha, Yu-Xiang Wang, Zhiqi Bu","submitted_at":"2022-09-30T18:38:53Z","abstract_excerpt":"Differentially private (DP) optimization is the standard paradigm to learn large neural networks that are accurate and privacy-preserving. The computational cost for DP deep learning, however, is notoriously heavy due to the per-sample gradient clipping. Existing DP implementations are 2-1000X more costly in time and space complexity than the standard (non-private) training. In this work, we develop a novel Book-Keeping (BK) technique that implements existing DP optimizers (thus achieving the same accuracy), with a substantial improvement on the computational cost. Specifically, BK enables DP "},"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":"2210.00038","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-09-30T18:38:53Z","cross_cats_sorted":["cs.CL","cs.CR","cs.CV"],"title_canon_sha256":"c8e503561a61632ab18ebbe2760d1ca34a8457a1433ba10d2997a37710f964da","abstract_canon_sha256":"fe310afbcf567d0d78160c46666f05cf85eb9fea06bc301cedc42b32e931c940"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:52:00.731296Z","signature_b64":"mBYDrrClqMv8dCWptg7FHvSlVvZ+nNjgD8Wd9cqPs8WQu8kK7Ke8dgvSVjMf12k2vMl/p4UQSP0iYBwAUPZ8Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7b89e8ec2cba2172d5ed63fd423ecac9e5ccecbd5d852ce8cdbe77d10d74d5dc","last_reissued_at":"2026-07-05T06:52:00.730758Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:52:00.730758Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Differentially Private Optimization on Large Model at Small Cost","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CR","cs.CV"],"primary_cat":"cs.LG","authors_text":"George Karypis, Sheng Zha, Yu-Xiang Wang, Zhiqi Bu","submitted_at":"2022-09-30T18:38:53Z","abstract_excerpt":"Differentially private (DP) optimization is the standard paradigm to learn large neural networks that are accurate and privacy-preserving. The computational cost for DP deep learning, however, is notoriously heavy due to the per-sample gradient clipping. Existing DP implementations are 2-1000X more costly in time and space complexity than the standard (non-private) training. In this work, we develop a novel Book-Keeping (BK) technique that implements existing DP optimizers (thus achieving the same accuracy), with a substantial improvement on the computational cost. Specifically, BK enables DP "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2210.00038","kind":"arxiv","version":2},"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/2210.00038/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":"2210.00038","created_at":"2026-07-05T06:52:00.730813+00:00"},{"alias_kind":"arxiv_version","alias_value":"2210.00038v2","created_at":"2026-07-05T06:52:00.730813+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2210.00038","created_at":"2026-07-05T06:52:00.730813+00:00"},{"alias_kind":"pith_short_12","alias_value":"POE6R3BMXIQX","created_at":"2026-07-05T06:52:00.730813+00:00"},{"alias_kind":"pith_short_16","alias_value":"POE6R3BMXIQXFVPN","created_at":"2026-07-05T06:52:00.730813+00:00"},{"alias_kind":"pith_short_8","alias_value":"POE6R3BM","created_at":"2026-07-05T06:52:00.730813+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/POE6R3BMXIQXFVPNMP6UEPWKZH","json":"https://pith.science/pith/POE6R3BMXIQXFVPNMP6UEPWKZH.json","graph_json":"https://pith.science/api/pith-number/POE6R3BMXIQXFVPNMP6UEPWKZH/graph.json","events_json":"https://pith.science/api/pith-number/POE6R3BMXIQXFVPNMP6UEPWKZH/events.json","paper":"https://pith.science/paper/POE6R3BM"},"agent_actions":{"view_html":"https://pith.science/pith/POE6R3BMXIQXFVPNMP6UEPWKZH","download_json":"https://pith.science/pith/POE6R3BMXIQXFVPNMP6UEPWKZH.json","view_paper":"https://pith.science/paper/POE6R3BM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2210.00038&json=true","fetch_graph":"https://pith.science/api/pith-number/POE6R3BMXIQXFVPNMP6UEPWKZH/graph.json","fetch_events":"https://pith.science/api/pith-number/POE6R3BMXIQXFVPNMP6UEPWKZH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/POE6R3BMXIQXFVPNMP6UEPWKZH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/POE6R3BMXIQXFVPNMP6UEPWKZH/action/storage_attestation","attest_author":"https://pith.science/pith/POE6R3BMXIQXFVPNMP6UEPWKZH/action/author_attestation","sign_citation":"https://pith.science/pith/POE6R3BMXIQXFVPNMP6UEPWKZH/action/citation_signature","submit_replication":"https://pith.science/pith/POE6R3BMXIQXFVPNMP6UEPWKZH/action/replication_record"}},"created_at":"2026-07-05T06:52:00.730813+00:00","updated_at":"2026-07-05T06:52:00.730813+00:00"}