{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:7UTMVQGK7CXBR6LSUKGO67TGLZ","short_pith_number":"pith:7UTMVQGK","schema_version":"1.0","canonical_sha256":"fd26cac0caf8ae18f972a28cef7e665e50395e6affd648008a7a597cdff933b8","source":{"kind":"arxiv","id":"2210.00688","version":3},"attestation_state":"computed","paper":{"title":"On the infinite-depth limit of finite-width neural networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","math.PR"],"primary_cat":"stat.ML","authors_text":"Soufiane Hayou","submitted_at":"2022-10-03T02:35:23Z","abstract_excerpt":"In this paper, we study the infinite-depth limit of finite-width residual neural networks with random Gaussian weights. With proper scaling, we show that by fixing the width and taking the depth to infinity, the pre-activations converge in distribution to a zero-drift diffusion process. Unlike the infinite-width limit where the pre-activation converge weakly to a Gaussian random variable, we show that the infinite-depth limit yields different distributions depending on the choice of the activation function. We document two cases where these distributions have closed-form (different) expression"},"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":"2210.00688","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2022-10-03T02:35:23Z","cross_cats_sorted":["cs.LG","math.PR"],"title_canon_sha256":"c6790e4f46f9e6d2ac4918fa9d4dc779cbaad3d7915c672e2d2439df1f1be409","abstract_canon_sha256":"633b5d3ac1e36d268009a3d070c6e7c179e2bae9ae9abb03ec225f7ab08d0c2b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:32:48.537232Z","signature_b64":"YgZY3P1SF0RfyH/mHbvSSZHxPgbe7QLsFV5me6vdoSInOrJYAxG4bNi0Xafvmxdpl1bmvHindPCzGm/sqfEjDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fd26cac0caf8ae18f972a28cef7e665e50395e6affd648008a7a597cdff933b8","last_reissued_at":"2026-07-05T05:32:48.536813Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:32:48.536813Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the infinite-depth limit of finite-width neural networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","math.PR"],"primary_cat":"stat.ML","authors_text":"Soufiane Hayou","submitted_at":"2022-10-03T02:35:23Z","abstract_excerpt":"In this paper, we study the infinite-depth limit of finite-width residual neural networks with random Gaussian weights. With proper scaling, we show that by fixing the width and taking the depth to infinity, the pre-activations converge in distribution to a zero-drift diffusion process. Unlike the infinite-width limit where the pre-activation converge weakly to a Gaussian random variable, we show that the infinite-depth limit yields different distributions depending on the choice of the activation function. We document two cases where these distributions have closed-form (different) expression"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2210.00688","kind":"arxiv","version":3},"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/2210.00688/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":"2210.00688","created_at":"2026-07-05T05:32:48.536869+00:00"},{"alias_kind":"arxiv_version","alias_value":"2210.00688v3","created_at":"2026-07-05T05:32:48.536869+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2210.00688","created_at":"2026-07-05T05:32:48.536869+00:00"},{"alias_kind":"pith_short_12","alias_value":"7UTMVQGK7CXB","created_at":"2026-07-05T05:32:48.536869+00:00"},{"alias_kind":"pith_short_16","alias_value":"7UTMVQGK7CXBR6LS","created_at":"2026-07-05T05:32:48.536869+00:00"},{"alias_kind":"pith_short_8","alias_value":"7UTMVQGK","created_at":"2026-07-05T05:32:48.536869+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.08218","citing_title":"How Deep Are Deep GPs, Really? A Sharp Threshold and a Non-Gaussian Limit for Compositional GPs","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2407.01602","citing_title":"Clustering in pure-attention hardmax transformers and its role in sentiment analysis","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7UTMVQGK7CXBR6LSUKGO67TGLZ","json":"https://pith.science/pith/7UTMVQGK7CXBR6LSUKGO67TGLZ.json","graph_json":"https://pith.science/api/pith-number/7UTMVQGK7CXBR6LSUKGO67TGLZ/graph.json","events_json":"https://pith.science/api/pith-number/7UTMVQGK7CXBR6LSUKGO67TGLZ/events.json","paper":"https://pith.science/paper/7UTMVQGK"},"agent_actions":{"view_html":"https://pith.science/pith/7UTMVQGK7CXBR6LSUKGO67TGLZ","download_json":"https://pith.science/pith/7UTMVQGK7CXBR6LSUKGO67TGLZ.json","view_paper":"https://pith.science/paper/7UTMVQGK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2210.00688&json=true","fetch_graph":"https://pith.science/api/pith-number/7UTMVQGK7CXBR6LSUKGO67TGLZ/graph.json","fetch_events":"https://pith.science/api/pith-number/7UTMVQGK7CXBR6LSUKGO67TGLZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7UTMVQGK7CXBR6LSUKGO67TGLZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7UTMVQGK7CXBR6LSUKGO67TGLZ/action/storage_attestation","attest_author":"https://pith.science/pith/7UTMVQGK7CXBR6LSUKGO67TGLZ/action/author_attestation","sign_citation":"https://pith.science/pith/7UTMVQGK7CXBR6LSUKGO67TGLZ/action/citation_signature","submit_replication":"https://pith.science/pith/7UTMVQGK7CXBR6LSUKGO67TGLZ/action/replication_record"}},"created_at":"2026-07-05T05:32:48.536869+00:00","updated_at":"2026-07-05T05:32:48.536869+00:00"}