{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:CB77XFHP7YBWGETYFU443TT4YE","short_pith_number":"pith:CB77XFHP","schema_version":"1.0","canonical_sha256":"107ffb94effe036312782d39cdce7cc1066d7ff5bbb19c74f5879f7095016c5a","source":{"kind":"arxiv","id":"2201.13011","version":2},"attestation_state":"computed","paper":{"title":"On the Power-Law Hessian Spectrums in Deep Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["physics.bio-ph","q-bio.BM"],"primary_cat":"cs.LG","authors_text":"Mingming Sun, Ping Li, Qian-Yuan Tang, Yunfeng Cai, Zeke Xie","submitted_at":"2022-01-31T06:04:47Z","abstract_excerpt":"It is well-known that the Hessian of deep loss landscape matters to optimization, generalization, and even robustness of deep learning. Recent works empirically discovered that the Hessian spectrum in deep learning has a two-component structure that consists of a small number of large eigenvalues and a large number of nearly-zero eigenvalues. However, the theoretical mechanism or the mathematical behind the Hessian spectrum is still largely under-explored. To the best of our knowledge, we are the first to demonstrate that the Hessian spectrums of well-trained deep neural networks exhibit simpl"},"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":"2201.13011","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-01-31T06:04:47Z","cross_cats_sorted":["physics.bio-ph","q-bio.BM"],"title_canon_sha256":"0728ae1150bbb34ff18d576312732ab3a6f1e188d2050b56df6fb2dd495bbc33","abstract_canon_sha256":"b81a0773a3f6d97cae038666e1772900c5f4c69df123d66695c256dcf53dc6f8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:45:00.147892Z","signature_b64":"8lD9aCJV1ZfQaolL7002yMZWXtzwdAHGn5J+UE1pZFgJbkoRLPeClcFShqV3i7Erew3eTTGdetK3tNSH/HE/Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"107ffb94effe036312782d39cdce7cc1066d7ff5bbb19c74f5879f7095016c5a","last_reissued_at":"2026-07-05T04:45:00.147414Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:45:00.147414Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the Power-Law Hessian Spectrums in Deep Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["physics.bio-ph","q-bio.BM"],"primary_cat":"cs.LG","authors_text":"Mingming Sun, Ping Li, Qian-Yuan Tang, Yunfeng Cai, Zeke Xie","submitted_at":"2022-01-31T06:04:47Z","abstract_excerpt":"It is well-known that the Hessian of deep loss landscape matters to optimization, generalization, and even robustness of deep learning. Recent works empirically discovered that the Hessian spectrum in deep learning has a two-component structure that consists of a small number of large eigenvalues and a large number of nearly-zero eigenvalues. However, the theoretical mechanism or the mathematical behind the Hessian spectrum is still largely under-explored. To the best of our knowledge, we are the first to demonstrate that the Hessian spectrums of well-trained deep neural networks exhibit simpl"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2201.13011","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/2201.13011/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":"2201.13011","created_at":"2026-07-05T04:45:00.147468+00:00"},{"alias_kind":"arxiv_version","alias_value":"2201.13011v2","created_at":"2026-07-05T04:45:00.147468+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2201.13011","created_at":"2026-07-05T04:45:00.147468+00:00"},{"alias_kind":"pith_short_12","alias_value":"CB77XFHP7YBW","created_at":"2026-07-05T04:45:00.147468+00:00"},{"alias_kind":"pith_short_16","alias_value":"CB77XFHP7YBWGETY","created_at":"2026-07-05T04:45:00.147468+00:00"},{"alias_kind":"pith_short_8","alias_value":"CB77XFHP","created_at":"2026-07-05T04:45:00.147468+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.28662","citing_title":"Closed-Form Steepest Descent Direction toward Flat Minima: Reducing Upper Bounds on the Loss Hessian Eigenspectrum in Neural Networks","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30619","citing_title":"Why can genetic algorithms work in high-dimensional search spaces?","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06081","citing_title":"Fast Gauss-Newton for Multiclass Cross-Entropy","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10202","citing_title":"Wolkowicz-Styan Upper Bound on the Hessian Eigenspectrum for Cross-Entropy Loss in Nonlinear Smooth Neural Networks","ref_index":48,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CB77XFHP7YBWGETYFU443TT4YE","json":"https://pith.science/pith/CB77XFHP7YBWGETYFU443TT4YE.json","graph_json":"https://pith.science/api/pith-number/CB77XFHP7YBWGETYFU443TT4YE/graph.json","events_json":"https://pith.science/api/pith-number/CB77XFHP7YBWGETYFU443TT4YE/events.json","paper":"https://pith.science/paper/CB77XFHP"},"agent_actions":{"view_html":"https://pith.science/pith/CB77XFHP7YBWGETYFU443TT4YE","download_json":"https://pith.science/pith/CB77XFHP7YBWGETYFU443TT4YE.json","view_paper":"https://pith.science/paper/CB77XFHP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2201.13011&json=true","fetch_graph":"https://pith.science/api/pith-number/CB77XFHP7YBWGETYFU443TT4YE/graph.json","fetch_events":"https://pith.science/api/pith-number/CB77XFHP7YBWGETYFU443TT4YE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CB77XFHP7YBWGETYFU443TT4YE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CB77XFHP7YBWGETYFU443TT4YE/action/storage_attestation","attest_author":"https://pith.science/pith/CB77XFHP7YBWGETYFU443TT4YE/action/author_attestation","sign_citation":"https://pith.science/pith/CB77XFHP7YBWGETYFU443TT4YE/action/citation_signature","submit_replication":"https://pith.science/pith/CB77XFHP7YBWGETYFU443TT4YE/action/replication_record"}},"created_at":"2026-07-05T04:45:00.147468+00:00","updated_at":"2026-07-05T04:45:00.147468+00:00"}