{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:HXZA4IWWF3OU3JPGYIZ4IBEXEN","short_pith_number":"pith:HXZA4IWW","schema_version":"1.0","canonical_sha256":"3df20e22d62edd4da5e6c233c404972366f183560e82a22d3f3e97bdc0335fd9","source":{"kind":"arxiv","id":"1805.11604","version":5},"attestation_state":"computed","paper":{"title":"How Does Batch Normalization Help Optimization?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","cs.NE"],"primary_cat":"stat.ML","authors_text":"Aleksander Madry, Andrew Ilyas, Dimitris Tsipras, Shibani Santurkar","submitted_at":"2018-05-29T17:42:00Z","abstract_excerpt":"Batch Normalization (BatchNorm) is a widely adopted technique that enables faster and more stable training of deep neural networks (DNNs). Despite its pervasiveness, the exact reasons for BatchNorm's effectiveness are still poorly understood. The popular belief is that this effectiveness stems from controlling the change of the layers' input distributions during training to reduce the so-called \"internal covariate shift\". In this work, we demonstrate that such distributional stability of layer inputs has little to do with the success of BatchNorm. Instead, we uncover a more fundamental impact "},"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":"1805.11604","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2018-05-29T17:42:00Z","cross_cats_sorted":["cs.LG","cs.NE"],"title_canon_sha256":"96aa45a17c11332d77cd3eab3cf669d7a9c3c69059fa46f1652db70973fb6953","abstract_canon_sha256":"364db1ad16be84982af643739b5a47d60141ccaf015ac360067f5712dafe8e67"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:48:40.456684Z","signature_b64":"LnjR7gBCcf5EMWLy1SXl21/7j6Q1GtK5VBUpsa910BEm1V9LSegycdaZk+gnD2GvI6jX6jWaIgYDJhonKNgCAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3df20e22d62edd4da5e6c233c404972366f183560e82a22d3f3e97bdc0335fd9","last_reissued_at":"2026-05-17T23:48:40.456057Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:48:40.456057Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Does Batch Normalization Help Optimization?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","cs.NE"],"primary_cat":"stat.ML","authors_text":"Aleksander Madry, Andrew Ilyas, Dimitris Tsipras, Shibani Santurkar","submitted_at":"2018-05-29T17:42:00Z","abstract_excerpt":"Batch Normalization (BatchNorm) is a widely adopted technique that enables faster and more stable training of deep neural networks (DNNs). Despite its pervasiveness, the exact reasons for BatchNorm's effectiveness are still poorly understood. The popular belief is that this effectiveness stems from controlling the change of the layers' input distributions during training to reduce the so-called \"internal covariate shift\". In this work, we demonstrate that such distributional stability of layer inputs has little to do with the success of BatchNorm. Instead, we uncover a more fundamental impact "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1805.11604","kind":"arxiv","version":5},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"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":"1805.11604","created_at":"2026-05-17T23:48:40.456159+00:00"},{"alias_kind":"arxiv_version","alias_value":"1805.11604v5","created_at":"2026-05-17T23:48:40.456159+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1805.11604","created_at":"2026-05-17T23:48:40.456159+00:00"},{"alias_kind":"pith_short_12","alias_value":"HXZA4IWWF3OU","created_at":"2026-05-18T12:32:28.185984+00:00"},{"alias_kind":"pith_short_16","alias_value":"HXZA4IWWF3OU3JPG","created_at":"2026-05-18T12:32:28.185984+00:00"},{"alias_kind":"pith_short_8","alias_value":"HXZA4IWW","created_at":"2026-05-18T12:32:28.185984+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"1906.10822","citing_title":"Gradient Noise Convolution (GNC): Smoothing Loss Function for Distributed Large-Batch SGD","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"1907.04003","citing_title":"Mean Spectral Normalization of Deep Neural Networks for Embedded Automation","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"1907.06286","citing_title":"Autoencoding sensory substitution","ref_index":209,"is_internal_anchor":true},{"citing_arxiv_id":"2104.13478","citing_title":"Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges","ref_index":75,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HXZA4IWWF3OU3JPGYIZ4IBEXEN","json":"https://pith.science/pith/HXZA4IWWF3OU3JPGYIZ4IBEXEN.json","graph_json":"https://pith.science/api/pith-number/HXZA4IWWF3OU3JPGYIZ4IBEXEN/graph.json","events_json":"https://pith.science/api/pith-number/HXZA4IWWF3OU3JPGYIZ4IBEXEN/events.json","paper":"https://pith.science/paper/HXZA4IWW"},"agent_actions":{"view_html":"https://pith.science/pith/HXZA4IWWF3OU3JPGYIZ4IBEXEN","download_json":"https://pith.science/pith/HXZA4IWWF3OU3JPGYIZ4IBEXEN.json","view_paper":"https://pith.science/paper/HXZA4IWW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1805.11604&json=true","fetch_graph":"https://pith.science/api/pith-number/HXZA4IWWF3OU3JPGYIZ4IBEXEN/graph.json","fetch_events":"https://pith.science/api/pith-number/HXZA4IWWF3OU3JPGYIZ4IBEXEN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HXZA4IWWF3OU3JPGYIZ4IBEXEN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HXZA4IWWF3OU3JPGYIZ4IBEXEN/action/storage_attestation","attest_author":"https://pith.science/pith/HXZA4IWWF3OU3JPGYIZ4IBEXEN/action/author_attestation","sign_citation":"https://pith.science/pith/HXZA4IWWF3OU3JPGYIZ4IBEXEN/action/citation_signature","submit_replication":"https://pith.science/pith/HXZA4IWWF3OU3JPGYIZ4IBEXEN/action/replication_record"}},"created_at":"2026-05-17T23:48:40.456159+00:00","updated_at":"2026-05-17T23:48:40.456159+00:00"}