{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:K4FQ62ZXWFXC6MW346A2TSLTHB","short_pith_number":"pith:K4FQ62ZX","schema_version":"1.0","canonical_sha256":"570b0f6b37b16e2f32dbe781a9c973384f31933752a45523fbf97fec9166e3ab","source":{"kind":"arxiv","id":"1910.05929","version":1},"attestation_state":"computed","paper":{"title":"Emergent properties of the local geometry of neural loss landscapes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.NE","stat.ML"],"primary_cat":"cs.LG","authors_text":"Stanislav Fort, Surya Ganguli","submitted_at":"2019-10-14T06:23:38Z","abstract_excerpt":"The local geometry of high dimensional neural network loss landscapes can both challenge our cherished theoretical intuitions as well as dramatically impact the practical success of neural network training. Indeed recent works have observed 4 striking local properties of neural loss landscapes on classification tasks: (1) the landscape exhibits exactly $C$ directions of high positive curvature, where $C$ is the number of classes; (2) gradient directions are largely confined to this extremely low dimensional subspace of positive Hessian curvature, leaving the vast majority of directions in weig"},"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":"1910.05929","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-10-14T06:23:38Z","cross_cats_sorted":["cs.NE","stat.ML"],"title_canon_sha256":"e7a7a10a440a43be1b2a38c109fda8c6fd34d8ef21532e5a024bb4793b5e6ef5","abstract_canon_sha256":"317a3d7d5e73d452972ef2847362e3757cd0c32b74dfa3dbcb61630012235d37"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T00:11:42.669570Z","signature_b64":"lzUKRwnWN9ZevRvRIdHOedcjjPI/2UAyE4oFbeh7V4EMBH4kK72KbnUVC7+7oBMookjQ4/2ok3TMxR0GynqkBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"570b0f6b37b16e2f32dbe781a9c973384f31933752a45523fbf97fec9166e3ab","last_reissued_at":"2026-07-05T00:11:42.669113Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T00:11:42.669113Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Emergent properties of the local geometry of neural loss landscapes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.NE","stat.ML"],"primary_cat":"cs.LG","authors_text":"Stanislav Fort, Surya Ganguli","submitted_at":"2019-10-14T06:23:38Z","abstract_excerpt":"The local geometry of high dimensional neural network loss landscapes can both challenge our cherished theoretical intuitions as well as dramatically impact the practical success of neural network training. Indeed recent works have observed 4 striking local properties of neural loss landscapes on classification tasks: (1) the landscape exhibits exactly $C$ directions of high positive curvature, where $C$ is the number of classes; (2) gradient directions are largely confined to this extremely low dimensional subspace of positive Hessian curvature, leaving the vast majority of directions in weig"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1910.05929","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/1910.05929/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":"1910.05929","created_at":"2026-07-05T00:11:42.669171+00:00"},{"alias_kind":"arxiv_version","alias_value":"1910.05929v1","created_at":"2026-07-05T00:11:42.669171+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1910.05929","created_at":"2026-07-05T00:11:42.669171+00:00"},{"alias_kind":"pith_short_12","alias_value":"K4FQ62ZXWFXC","created_at":"2026-07-05T00:11:42.669171+00:00"},{"alias_kind":"pith_short_16","alias_value":"K4FQ62ZXWFXC6MW3","created_at":"2026-07-05T00:11:42.669171+00:00"},{"alias_kind":"pith_short_8","alias_value":"K4FQ62ZX","created_at":"2026-07-05T00:11:42.669171+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07845","citing_title":"Explaining Near-Zero Hessian Eigenvalues Through Approximate Symmetries in Neural Networks","ref_index":36,"is_internal_anchor":true},{"citing_arxiv_id":"2606.13637","citing_title":"The Stable Recovery Manifold: Geometric Principles Governing Recoverability in Continual Learning","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04662","citing_title":"Why Muon Outperforms Adam: A Curvature Perspective","ref_index":141,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30226","citing_title":"Characterizing Optimizer-Dependent Training Dynamics Through Hessian Eigenvector Displacement and Localization","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28986","citing_title":"Comparing Classical Simulation and Sample-Based Learning of Quantum Systems","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07790","citing_title":"Hessian Surgery: Class-Targeted Post-Hoc Rebalancing via Hessian Spike Perturbation","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07790","citing_title":"Hessian Surgery: Class-Targeted Post-Hoc Rebalancing via Hessian Spike Perturbation","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16431","citing_title":"Dimensional Criticality at Grokking Across MLPs and Transformers","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04655","citing_title":"Grokking as Dimensional Phase Transition in Neural Networks","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/K4FQ62ZXWFXC6MW346A2TSLTHB","json":"https://pith.science/pith/K4FQ62ZXWFXC6MW346A2TSLTHB.json","graph_json":"https://pith.science/api/pith-number/K4FQ62ZXWFXC6MW346A2TSLTHB/graph.json","events_json":"https://pith.science/api/pith-number/K4FQ62ZXWFXC6MW346A2TSLTHB/events.json","paper":"https://pith.science/paper/K4FQ62ZX"},"agent_actions":{"view_html":"https://pith.science/pith/K4FQ62ZXWFXC6MW346A2TSLTHB","download_json":"https://pith.science/pith/K4FQ62ZXWFXC6MW346A2TSLTHB.json","view_paper":"https://pith.science/paper/K4FQ62ZX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1910.05929&json=true","fetch_graph":"https://pith.science/api/pith-number/K4FQ62ZXWFXC6MW346A2TSLTHB/graph.json","fetch_events":"https://pith.science/api/pith-number/K4FQ62ZXWFXC6MW346A2TSLTHB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/K4FQ62ZXWFXC6MW346A2TSLTHB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/K4FQ62ZXWFXC6MW346A2TSLTHB/action/storage_attestation","attest_author":"https://pith.science/pith/K4FQ62ZXWFXC6MW346A2TSLTHB/action/author_attestation","sign_citation":"https://pith.science/pith/K4FQ62ZXWFXC6MW346A2TSLTHB/action/citation_signature","submit_replication":"https://pith.science/pith/K4FQ62ZXWFXC6MW346A2TSLTHB/action/replication_record"}},"created_at":"2026-07-05T00:11:42.669171+00:00","updated_at":"2026-07-05T00:11:42.669171+00:00"}