{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:5TKJMTWNK3RYYH25MATZQKCC35","short_pith_number":"pith:5TKJMTWN","schema_version":"1.0","canonical_sha256":"ecd4964ecd56e38c1f5d6027982842df7f9c79657cd6acb5031ad96c2104e4a5","source":{"kind":"arxiv","id":"2003.00152","version":3},"attestation_state":"computed","paper":{"title":"Training BatchNorm and Only BatchNorm: On the Expressive Power of Random Features in CNNs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.NE","stat.ML"],"primary_cat":"cs.LG","authors_text":"Ari S. Morcos, David J. Schwab, Jonathan Frankle","submitted_at":"2020-02-29T01:57:37Z","abstract_excerpt":"A wide variety of deep learning techniques from style transfer to multitask learning rely on training affine transformations of features. Most prominent among these is the popular feature normalization technique BatchNorm, which normalizes activations and then subsequently applies a learned affine transform. In this paper, we aim to understand the role and expressive power of affine parameters used to transform features in this way. To isolate the contribution of these parameters from that of the learned features they transform, we investigate the performance achieved when training only these "},"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":"2003.00152","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-02-29T01:57:37Z","cross_cats_sorted":["cs.AI","cs.NE","stat.ML"],"title_canon_sha256":"a7c7d20b11044574be43ebcf8e928ae4a51150598b65f24b581d55db0e75c7f2","abstract_canon_sha256":"3c9eb69a89b4a577269acb7add3839f71a22ea65aa52320bd0a6050670df47c2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:24:48.846195Z","signature_b64":"mFACdQGSM/SyQGJWPPqcYOqHKu2tJb7bfxCcvd7rvznr+X2x8SQpq0OPPv8w7vHe/4twkW7I+bqDwXeEIvZmCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ecd4964ecd56e38c1f5d6027982842df7f9c79657cd6acb5031ad96c2104e4a5","last_reissued_at":"2026-07-05T02:24:48.845734Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:24:48.845734Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Training BatchNorm and Only BatchNorm: On the Expressive Power of Random Features in CNNs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.NE","stat.ML"],"primary_cat":"cs.LG","authors_text":"Ari S. Morcos, David J. Schwab, Jonathan Frankle","submitted_at":"2020-02-29T01:57:37Z","abstract_excerpt":"A wide variety of deep learning techniques from style transfer to multitask learning rely on training affine transformations of features. Most prominent among these is the popular feature normalization technique BatchNorm, which normalizes activations and then subsequently applies a learned affine transform. In this paper, we aim to understand the role and expressive power of affine parameters used to transform features in this way. To isolate the contribution of these parameters from that of the learned features they transform, we investigate the performance achieved when training only these "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2003.00152","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/2003.00152/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":"2003.00152","created_at":"2026-07-05T02:24:48.845808+00:00"},{"alias_kind":"arxiv_version","alias_value":"2003.00152v3","created_at":"2026-07-05T02:24:48.845808+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2003.00152","created_at":"2026-07-05T02:24:48.845808+00:00"},{"alias_kind":"pith_short_12","alias_value":"5TKJMTWNK3RY","created_at":"2026-07-05T02:24:48.845808+00:00"},{"alias_kind":"pith_short_16","alias_value":"5TKJMTWNK3RYYH25","created_at":"2026-07-05T02:24:48.845808+00:00"},{"alias_kind":"pith_short_8","alias_value":"5TKJMTWN","created_at":"2026-07-05T02:24:48.845808+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2508.15452","citing_title":"DoSReMC: Domain Shift Resilient Mammography Classification using Batch Normalization Adaptation","ref_index":59,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5TKJMTWNK3RYYH25MATZQKCC35","json":"https://pith.science/pith/5TKJMTWNK3RYYH25MATZQKCC35.json","graph_json":"https://pith.science/api/pith-number/5TKJMTWNK3RYYH25MATZQKCC35/graph.json","events_json":"https://pith.science/api/pith-number/5TKJMTWNK3RYYH25MATZQKCC35/events.json","paper":"https://pith.science/paper/5TKJMTWN"},"agent_actions":{"view_html":"https://pith.science/pith/5TKJMTWNK3RYYH25MATZQKCC35","download_json":"https://pith.science/pith/5TKJMTWNK3RYYH25MATZQKCC35.json","view_paper":"https://pith.science/paper/5TKJMTWN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2003.00152&json=true","fetch_graph":"https://pith.science/api/pith-number/5TKJMTWNK3RYYH25MATZQKCC35/graph.json","fetch_events":"https://pith.science/api/pith-number/5TKJMTWNK3RYYH25MATZQKCC35/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5TKJMTWNK3RYYH25MATZQKCC35/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5TKJMTWNK3RYYH25MATZQKCC35/action/storage_attestation","attest_author":"https://pith.science/pith/5TKJMTWNK3RYYH25MATZQKCC35/action/author_attestation","sign_citation":"https://pith.science/pith/5TKJMTWNK3RYYH25MATZQKCC35/action/citation_signature","submit_replication":"https://pith.science/pith/5TKJMTWNK3RYYH25MATZQKCC35/action/replication_record"}},"created_at":"2026-07-05T02:24:48.845808+00:00","updated_at":"2026-07-05T02:24:48.845808+00:00"}