{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:V4EJ7K47FMBB4SDMPBGDWWTKFZ","short_pith_number":"pith:V4EJ7K47","schema_version":"1.0","canonical_sha256":"af089fab9f2b021e486c784c3b5a6a2e5058ed90d732054a891a36e667847e46","source":{"kind":"arxiv","id":"2605.20450","version":1},"attestation_state":"computed","paper":{"title":"SMA-DP: Spectral Memory-Aware Differential Privacy for Deep Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Mohammad Partohaghighi, Roummel Marcia","submitted_at":"2026-05-19T20:02:19Z","abstract_excerpt":"Differentially private stochastic gradient descent (DP-SGD) enables private deep learning through per-example clipping and calibrated Gaussian noise, but its high-variance updates can reduce utility on challenging datasets. We propose \\textbf{SMA-DP-SGD}, a \\textbf{Spectral Memory-Aware Differentially Private Stochastic Gradient Descent} method that augments DP-SGD with a fractional memory branch built only from previously privatized noisy releases. WeightWatcher-inspired power-law spectral exponents provide group-wise reliability signals, instantiated layer-wise in our experiments, to adapt t"},"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":"2605.20450","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-19T20:02:19Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"89eedff697b1efb0cd5425a55a6d08d37156fea5d12200bc203ff6bcb98a2bdc","abstract_canon_sha256":"8b19b11c32fb5498fcb8cb29be21340c73054efba7698111705afa627329dbe1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-21T01:04:38.406742Z","signature_b64":"d6TCY9VUuqXF9eqK43O4OfL046jKzw4UvGKDtyzH2AlgeM3wSpnbnWc8fTWgyr+4k2jVvKNmVqN2LFTMwfgaBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"af089fab9f2b021e486c784c3b5a6a2e5058ed90d732054a891a36e667847e46","last_reissued_at":"2026-05-21T01:04:38.405932Z","signature_status":"signed_v1","first_computed_at":"2026-05-21T01:04:38.405932Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SMA-DP: Spectral Memory-Aware Differential Privacy for Deep Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.LG","authors_text":"Mohammad Partohaghighi, Roummel Marcia","submitted_at":"2026-05-19T20:02:19Z","abstract_excerpt":"Differentially private stochastic gradient descent (DP-SGD) enables private deep learning through per-example clipping and calibrated Gaussian noise, but its high-variance updates can reduce utility on challenging datasets. We propose \\textbf{SMA-DP-SGD}, a \\textbf{Spectral Memory-Aware Differentially Private Stochastic Gradient Descent} method that augments DP-SGD with a fractional memory branch built only from previously privatized noisy releases. WeightWatcher-inspired power-law spectral exponents provide group-wise reliability signals, instantiated layer-wise in our experiments, to adapt t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.20450","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/2605.20450/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":"2605.20450","created_at":"2026-05-21T01:04:38.406043+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.20450v1","created_at":"2026-05-21T01:04:38.406043+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.20450","created_at":"2026-05-21T01:04:38.406043+00:00"},{"alias_kind":"pith_short_12","alias_value":"V4EJ7K47FMBB","created_at":"2026-05-21T01:04:38.406043+00:00"},{"alias_kind":"pith_short_16","alias_value":"V4EJ7K47FMBB4SDM","created_at":"2026-05-21T01:04:38.406043+00:00"},{"alias_kind":"pith_short_8","alias_value":"V4EJ7K47","created_at":"2026-05-21T01:04:38.406043+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/V4EJ7K47FMBB4SDMPBGDWWTKFZ","json":"https://pith.science/pith/V4EJ7K47FMBB4SDMPBGDWWTKFZ.json","graph_json":"https://pith.science/api/pith-number/V4EJ7K47FMBB4SDMPBGDWWTKFZ/graph.json","events_json":"https://pith.science/api/pith-number/V4EJ7K47FMBB4SDMPBGDWWTKFZ/events.json","paper":"https://pith.science/paper/V4EJ7K47"},"agent_actions":{"view_html":"https://pith.science/pith/V4EJ7K47FMBB4SDMPBGDWWTKFZ","download_json":"https://pith.science/pith/V4EJ7K47FMBB4SDMPBGDWWTKFZ.json","view_paper":"https://pith.science/paper/V4EJ7K47","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.20450&json=true","fetch_graph":"https://pith.science/api/pith-number/V4EJ7K47FMBB4SDMPBGDWWTKFZ/graph.json","fetch_events":"https://pith.science/api/pith-number/V4EJ7K47FMBB4SDMPBGDWWTKFZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V4EJ7K47FMBB4SDMPBGDWWTKFZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V4EJ7K47FMBB4SDMPBGDWWTKFZ/action/storage_attestation","attest_author":"https://pith.science/pith/V4EJ7K47FMBB4SDMPBGDWWTKFZ/action/author_attestation","sign_citation":"https://pith.science/pith/V4EJ7K47FMBB4SDMPBGDWWTKFZ/action/citation_signature","submit_replication":"https://pith.science/pith/V4EJ7K47FMBB4SDMPBGDWWTKFZ/action/replication_record"}},"created_at":"2026-05-21T01:04:38.406043+00:00","updated_at":"2026-05-21T01:04:38.406043+00:00"}