{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:HQEJJINJOUUMJOTJ3BGFEDMHX2","short_pith_number":"pith:HQEJJINJ","schema_version":"1.0","canonical_sha256":"3c0894a1a97528c4ba69d84c520d87beb46dcb44485275498eba9a911e2e321e","source":{"kind":"arxiv","id":"2106.10165","version":2},"attestation_state":"computed","paper":{"title":"The Principles of Deep Learning Theory","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","hep-th","stat.ML"],"primary_cat":"cs.LG","authors_text":"Boris Hanin, Daniel A. Roberts, Sho Yaida","submitted_at":"2021-06-18T15:00:00Z","abstract_excerpt":"This book develops an effective theory approach to understanding deep neural networks of practical relevance. Beginning from a first-principles component-level picture of networks, we explain how to determine an accurate description of the output of trained networks by solving layer-to-layer iteration equations and nonlinear learning dynamics. A main result is that the predictions of networks are described by nearly-Gaussian distributions, with the depth-to-width aspect ratio of the network controlling the deviations from the infinite-width Gaussian description. We explain how these effectivel"},"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":"2106.10165","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-18T15:00:00Z","cross_cats_sorted":["cs.AI","hep-th","stat.ML"],"title_canon_sha256":"34c82c7c22dad51704689b1ed256cab1edfe276ffcdf2980169701521b58eb2b","abstract_canon_sha256":"25ae3feff55aa3c7bc2e978769321d36e754681a89cf59d9cddbd13f883837cb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:27:06.108466Z","signature_b64":"BfXyFxp7ODbMyb6ts4qHlXAFE6Db9qok1OYda1nHOk1mqEUqxWDnREAUZOZ34fFk4Jx8EWUObOFSD8Cc6BUyDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3c0894a1a97528c4ba69d84c520d87beb46dcb44485275498eba9a911e2e321e","last_reissued_at":"2026-07-05T04:27:06.107993Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:27:06.107993Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Principles of Deep Learning Theory","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","hep-th","stat.ML"],"primary_cat":"cs.LG","authors_text":"Boris Hanin, Daniel A. Roberts, Sho Yaida","submitted_at":"2021-06-18T15:00:00Z","abstract_excerpt":"This book develops an effective theory approach to understanding deep neural networks of practical relevance. Beginning from a first-principles component-level picture of networks, we explain how to determine an accurate description of the output of trained networks by solving layer-to-layer iteration equations and nonlinear learning dynamics. A main result is that the predictions of networks are described by nearly-Gaussian distributions, with the depth-to-width aspect ratio of the network controlling the deviations from the infinite-width Gaussian description. We explain how these effectivel"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.10165","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/2106.10165/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":"2106.10165","created_at":"2026-07-05T04:27:06.108048+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.10165v2","created_at":"2026-07-05T04:27:06.108048+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.10165","created_at":"2026-07-05T04:27:06.108048+00:00"},{"alias_kind":"pith_short_12","alias_value":"HQEJJINJOUUM","created_at":"2026-07-05T04:27:06.108048+00:00"},{"alias_kind":"pith_short_16","alias_value":"HQEJJINJOUUMJOTJ","created_at":"2026-07-05T04:27:06.108048+00:00"},{"alias_kind":"pith_short_8","alias_value":"HQEJJINJ","created_at":"2026-07-05T04:27:06.108048+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11317","citing_title":"Lectures on Semiclassical Methods for Composite Operators","ref_index":118,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09950","citing_title":"Integrating Out, Twice:The Open-System Case That Neural-Network Ensemble Theory Is Missing","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2508.03810","citing_title":"Viability of perturbative expansion for quantum field theories on neurons","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2510.21742","citing_title":"Statistics of correlations in nonlinear recurrent neural networks","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27052","citing_title":"Man, Machine, and Mathematics","ref_index":78,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06563","citing_title":"Criticality and Saturation in Orthogonal Neural Networks","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HQEJJINJOUUMJOTJ3BGFEDMHX2","json":"https://pith.science/pith/HQEJJINJOUUMJOTJ3BGFEDMHX2.json","graph_json":"https://pith.science/api/pith-number/HQEJJINJOUUMJOTJ3BGFEDMHX2/graph.json","events_json":"https://pith.science/api/pith-number/HQEJJINJOUUMJOTJ3BGFEDMHX2/events.json","paper":"https://pith.science/paper/HQEJJINJ"},"agent_actions":{"view_html":"https://pith.science/pith/HQEJJINJOUUMJOTJ3BGFEDMHX2","download_json":"https://pith.science/pith/HQEJJINJOUUMJOTJ3BGFEDMHX2.json","view_paper":"https://pith.science/paper/HQEJJINJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.10165&json=true","fetch_graph":"https://pith.science/api/pith-number/HQEJJINJOUUMJOTJ3BGFEDMHX2/graph.json","fetch_events":"https://pith.science/api/pith-number/HQEJJINJOUUMJOTJ3BGFEDMHX2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HQEJJINJOUUMJOTJ3BGFEDMHX2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HQEJJINJOUUMJOTJ3BGFEDMHX2/action/storage_attestation","attest_author":"https://pith.science/pith/HQEJJINJOUUMJOTJ3BGFEDMHX2/action/author_attestation","sign_citation":"https://pith.science/pith/HQEJJINJOUUMJOTJ3BGFEDMHX2/action/citation_signature","submit_replication":"https://pith.science/pith/HQEJJINJOUUMJOTJ3BGFEDMHX2/action/replication_record"}},"created_at":"2026-07-05T04:27:06.108048+00:00","updated_at":"2026-07-05T04:27:06.108048+00:00"}