{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:2JZRKJAIVNTFVEDX3U4GYVFQCJ","short_pith_number":"pith:2JZRKJAI","schema_version":"1.0","canonical_sha256":"d273152408ab665a9077dd386c54b012690d2635aa82439e00df8c0f981d4b6d","source":{"kind":"arxiv","id":"1910.12744","version":2},"attestation_state":"computed","paper":{"title":"On approximating $\\nabla f$ with neural networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Saeed Saremi","submitted_at":"2019-10-28T15:12:42Z","abstract_excerpt":"Consider a feedforward neural network $\\psi: \\mathbb{R}^d\\rightarrow \\mathbb{R}^d$ such that $\\psi\\approx \\nabla f$, where $f:\\mathbb{R}^d \\rightarrow \\mathbb{R}$ is a smooth function, therefore $\\psi$ must satisfy $\\partial_j \\psi_i = \\partial_i \\psi_j$ pointwise. We prove a theorem that a $\\psi$ network with more than one hidden layer can only represent one feature in its first hidden layer; this is a dramatic departure from the well-known results for one hidden layer. The proof of the theorem is straightforward, where two backward paths and a weight-tying matrix play the key roles. We then "},"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":"1910.12744","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2019-10-28T15:12:42Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"73621ebd3d1ab802b62ba1d2481fa3c426049fbc63e38771a7983a0c18b56e3a","abstract_canon_sha256":"2d72ba50e27411fb4de8d2a506b44082b0508b7a0f9152737633e11556c65b52"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T00:17:29.549287Z","signature_b64":"KboOlFEqaYC6TEgaZSEvRPzWt99KapidwtOvlM42pkG/DsasQBgXsvRX+sS1UnU7PgFifFDkRcsT0/ceshebBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d273152408ab665a9077dd386c54b012690d2635aa82439e00df8c0f981d4b6d","last_reissued_at":"2026-07-05T00:17:29.548877Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T00:17:29.548877Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On approximating $\\nabla f$ with neural networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Saeed Saremi","submitted_at":"2019-10-28T15:12:42Z","abstract_excerpt":"Consider a feedforward neural network $\\psi: \\mathbb{R}^d\\rightarrow \\mathbb{R}^d$ such that $\\psi\\approx \\nabla f$, where $f:\\mathbb{R}^d \\rightarrow \\mathbb{R}$ is a smooth function, therefore $\\psi$ must satisfy $\\partial_j \\psi_i = \\partial_i \\psi_j$ pointwise. We prove a theorem that a $\\psi$ network with more than one hidden layer can only represent one feature in its first hidden layer; this is a dramatic departure from the well-known results for one hidden layer. The proof of the theorem is straightforward, where two backward paths and a weight-tying matrix play the key roles. We then "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1910.12744","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/1910.12744/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":"1910.12744","created_at":"2026-07-05T00:17:29.548933+00:00"},{"alias_kind":"arxiv_version","alias_value":"1910.12744v2","created_at":"2026-07-05T00:17:29.548933+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1910.12744","created_at":"2026-07-05T00:17:29.548933+00:00"},{"alias_kind":"pith_short_12","alias_value":"2JZRKJAIVNTF","created_at":"2026-07-05T00:17:29.548933+00:00"},{"alias_kind":"pith_short_16","alias_value":"2JZRKJAIVNTFVEDX","created_at":"2026-07-05T00:17:29.548933+00:00"},{"alias_kind":"pith_short_8","alias_value":"2JZRKJAI","created_at":"2026-07-05T00:17:29.548933+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.06179","citing_title":"Diffusion Models Observe Only Gradients: A Geometric Perspective on Score Matching Errors","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2509.04477","citing_title":"Universal Representation of Generalized Convex Functions and their Gradients","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07513","citing_title":"Tessellations of Semi-Discrete Flow Matching","ref_index":171,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2JZRKJAIVNTFVEDX3U4GYVFQCJ","json":"https://pith.science/pith/2JZRKJAIVNTFVEDX3U4GYVFQCJ.json","graph_json":"https://pith.science/api/pith-number/2JZRKJAIVNTFVEDX3U4GYVFQCJ/graph.json","events_json":"https://pith.science/api/pith-number/2JZRKJAIVNTFVEDX3U4GYVFQCJ/events.json","paper":"https://pith.science/paper/2JZRKJAI"},"agent_actions":{"view_html":"https://pith.science/pith/2JZRKJAIVNTFVEDX3U4GYVFQCJ","download_json":"https://pith.science/pith/2JZRKJAIVNTFVEDX3U4GYVFQCJ.json","view_paper":"https://pith.science/paper/2JZRKJAI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1910.12744&json=true","fetch_graph":"https://pith.science/api/pith-number/2JZRKJAIVNTFVEDX3U4GYVFQCJ/graph.json","fetch_events":"https://pith.science/api/pith-number/2JZRKJAIVNTFVEDX3U4GYVFQCJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2JZRKJAIVNTFVEDX3U4GYVFQCJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2JZRKJAIVNTFVEDX3U4GYVFQCJ/action/storage_attestation","attest_author":"https://pith.science/pith/2JZRKJAIVNTFVEDX3U4GYVFQCJ/action/author_attestation","sign_citation":"https://pith.science/pith/2JZRKJAIVNTFVEDX3U4GYVFQCJ/action/citation_signature","submit_replication":"https://pith.science/pith/2JZRKJAIVNTFVEDX3U4GYVFQCJ/action/replication_record"}},"created_at":"2026-07-05T00:17:29.548933+00:00","updated_at":"2026-07-05T00:17:29.548933+00:00"}