{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:63VM5QHQGJDSY7X5B7YT4YJ43B","short_pith_number":"pith:63VM5QHQ","schema_version":"1.0","canonical_sha256":"f6eacec0f032472c7efd0ff13e613cd8475770541190d6fe5e3150dd995b37fd","source":{"kind":"arxiv","id":"1902.06720","version":4},"attestation_state":"computed","paper":{"title":"Wide Neural Networks of Any Depth Evolve as Linear Models Under Gradient Descent","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Jaehoon Lee, Jascha Sohl-Dickstein, Jeffrey Pennington, Lechao Xiao, Roman Novak, Samuel S. Schoenholz, Yasaman Bahri","submitted_at":"2019-02-18T18:37:49Z","abstract_excerpt":"A longstanding goal in deep learning research has been to precisely characterize training and generalization. However, the often complex loss landscapes of neural networks have made a theory of learning dynamics elusive. In this work, we show that for wide neural networks the learning dynamics simplify considerably and that, in the infinite width limit, they are governed by a linear model obtained from the first-order Taylor expansion of the network around its initial parameters. Furthermore, mirroring the correspondence between wide Bayesian neural networks and Gaussian processes, gradient-ba"},"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":"1902.06720","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2019-02-18T18:37:49Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a3df410a2a8834ba4a89aa99418103bbcb3866e4c416aa2dade7f3f499309a16","abstract_canon_sha256":"ca740481f114c893cd0bf3134499dfb29264845d80ee2633da41565aad0757c5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:11:38.333651Z","signature_b64":"60q+EPH1e5zQ3mpJmJqM4RqTIZGN4kY9hzBo789/oB2+or1fRfadmkOWw7CbKxGZA2Zlzhcc3tv6QLfgstnMCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f6eacec0f032472c7efd0ff13e613cd8475770541190d6fe5e3150dd995b37fd","last_reissued_at":"2026-07-05T02:11:38.333224Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:11:38.333224Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Wide Neural Networks of Any Depth Evolve as Linear Models Under Gradient Descent","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Jaehoon Lee, Jascha Sohl-Dickstein, Jeffrey Pennington, Lechao Xiao, Roman Novak, Samuel S. Schoenholz, Yasaman Bahri","submitted_at":"2019-02-18T18:37:49Z","abstract_excerpt":"A longstanding goal in deep learning research has been to precisely characterize training and generalization. However, the often complex loss landscapes of neural networks have made a theory of learning dynamics elusive. In this work, we show that for wide neural networks the learning dynamics simplify considerably and that, in the infinite width limit, they are governed by a linear model obtained from the first-order Taylor expansion of the network around its initial parameters. Furthermore, mirroring the correspondence between wide Bayesian neural networks and Gaussian processes, gradient-ba"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1902.06720","kind":"arxiv","version":4},"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/1902.06720/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":"1902.06720","created_at":"2026-07-05T02:11:38.333284+00:00"},{"alias_kind":"arxiv_version","alias_value":"1902.06720v4","created_at":"2026-07-05T02:11:38.333284+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1902.06720","created_at":"2026-07-05T02:11:38.333284+00:00"},{"alias_kind":"pith_short_12","alias_value":"63VM5QHQGJDS","created_at":"2026-07-05T02:11:38.333284+00:00"},{"alias_kind":"pith_short_16","alias_value":"63VM5QHQGJDSY7X5","created_at":"2026-07-05T02:11:38.333284+00:00"},{"alias_kind":"pith_short_8","alias_value":"63VM5QHQ","created_at":"2026-07-05T02:11:38.333284+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.08316","citing_title":"Some Inverse Problems in Particle Physics","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30831","citing_title":"Geometric Dyson Brownian Motions and the Free Log-Normal Limit for a Non-Square Gaussian Matrix Product","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28585","citing_title":"Outer-Momentum Restarting in High-Dimensional Two-Phase Optimization","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"1906.08654","citing_title":"ID3 Learns Juntas for Smoothed Product Distributions","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"1906.08899","citing_title":"Limitations of Lazy Training of Two-layers Neural Networks","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2010.14701","citing_title":"Scaling Laws for Autoregressive Generative Modeling","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2001.08361","citing_title":"Scaling Laws for Neural Language Models","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/63VM5QHQGJDSY7X5B7YT4YJ43B","json":"https://pith.science/pith/63VM5QHQGJDSY7X5B7YT4YJ43B.json","graph_json":"https://pith.science/api/pith-number/63VM5QHQGJDSY7X5B7YT4YJ43B/graph.json","events_json":"https://pith.science/api/pith-number/63VM5QHQGJDSY7X5B7YT4YJ43B/events.json","paper":"https://pith.science/paper/63VM5QHQ"},"agent_actions":{"view_html":"https://pith.science/pith/63VM5QHQGJDSY7X5B7YT4YJ43B","download_json":"https://pith.science/pith/63VM5QHQGJDSY7X5B7YT4YJ43B.json","view_paper":"https://pith.science/paper/63VM5QHQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1902.06720&json=true","fetch_graph":"https://pith.science/api/pith-number/63VM5QHQGJDSY7X5B7YT4YJ43B/graph.json","fetch_events":"https://pith.science/api/pith-number/63VM5QHQGJDSY7X5B7YT4YJ43B/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/63VM5QHQGJDSY7X5B7YT4YJ43B/action/timestamp_anchor","attest_storage":"https://pith.science/pith/63VM5QHQGJDSY7X5B7YT4YJ43B/action/storage_attestation","attest_author":"https://pith.science/pith/63VM5QHQGJDSY7X5B7YT4YJ43B/action/author_attestation","sign_citation":"https://pith.science/pith/63VM5QHQGJDSY7X5B7YT4YJ43B/action/citation_signature","submit_replication":"https://pith.science/pith/63VM5QHQGJDSY7X5B7YT4YJ43B/action/replication_record"}},"created_at":"2026-07-05T02:11:38.333284+00:00","updated_at":"2026-07-05T02:11:38.333284+00:00"}