{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:D72DBSIFWZEEFOMZE4IHZVEEBT","short_pith_number":"pith:D72DBSIF","schema_version":"1.0","canonical_sha256":"1ff430c905b64842b99927107cd4840ce05e6dc544310602d9bcb2def4fcdef4","source":{"kind":"arxiv","id":"2106.00734","version":2},"attestation_state":"computed","paper":{"title":"Post-mortem on a deep learning contest: a Simpson's paradox and the complementary roles of scale metrics versus shape metrics","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Charles H. Martin, Michael W. Mahoney","submitted_at":"2021-06-01T19:19:49Z","abstract_excerpt":"To understand better good generalization performance in state-of-the-art neural network (NN) models, and in particular the success of the ALPHAHAT metric based on Heavy-Tailed Self-Regularization (HT-SR) theory, we analyze of a corpus of models that was made publicly-available for a contest to predict the generalization accuracy of NNs. These models include a wide range of qualities and were trained with a range of architectures and regularization hyperparameters. We break ALPHAHAT into its two subcomponent metrics: a scale-based metric; and a shape-based metric. We identify what amounts to a "},"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":"2106.00734","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-01T19:19:49Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"8fcd753ceb9f7d0a9e921476ae834570707779b381a413e3e510ed6a8f395218","abstract_canon_sha256":"0abd4520b6a8571f9b6a666b48cc7e344bbb549d81328ff35f094f1d24816588"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:54:41.441725Z","signature_b64":"nEy26bQH5rNd+TIdWigk5+gJlOAjNoD5biVQeIcBnJ8B3d4j1COvCEFgWn7bWJF9cLrvlHIsAai9xXSWXKUHCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1ff430c905b64842b99927107cd4840ce05e6dc544310602d9bcb2def4fcdef4","last_reissued_at":"2026-07-05T03:54:41.441323Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:54:41.441323Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Post-mortem on a deep learning contest: a Simpson's paradox and the complementary roles of scale metrics versus shape metrics","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Charles H. Martin, Michael W. Mahoney","submitted_at":"2021-06-01T19:19:49Z","abstract_excerpt":"To understand better good generalization performance in state-of-the-art neural network (NN) models, and in particular the success of the ALPHAHAT metric based on Heavy-Tailed Self-Regularization (HT-SR) theory, we analyze of a corpus of models that was made publicly-available for a contest to predict the generalization accuracy of NNs. These models include a wide range of qualities and were trained with a range of architectures and regularization hyperparameters. We break ALPHAHAT into its two subcomponent metrics: a scale-based metric; and a shape-based metric. We identify what amounts to a "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.00734","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/2106.00734/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":"2106.00734","created_at":"2026-07-05T03:54:41.441379+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.00734v2","created_at":"2026-07-05T03:54:41.441379+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.00734","created_at":"2026-07-05T03:54:41.441379+00:00"},{"alias_kind":"pith_short_12","alias_value":"D72DBSIFWZEE","created_at":"2026-07-05T03:54:41.441379+00:00"},{"alias_kind":"pith_short_16","alias_value":"D72DBSIFWZEEFOMZ","created_at":"2026-07-05T03:54:41.441379+00:00"},{"alias_kind":"pith_short_8","alias_value":"D72DBSIF","created_at":"2026-07-05T03:54:41.441379+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.17912","citing_title":"SETOL: A Semi-Empirical Theory of (Deep) Learning","ref_index":24,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/D72DBSIFWZEEFOMZE4IHZVEEBT","json":"https://pith.science/pith/D72DBSIFWZEEFOMZE4IHZVEEBT.json","graph_json":"https://pith.science/api/pith-number/D72DBSIFWZEEFOMZE4IHZVEEBT/graph.json","events_json":"https://pith.science/api/pith-number/D72DBSIFWZEEFOMZE4IHZVEEBT/events.json","paper":"https://pith.science/paper/D72DBSIF"},"agent_actions":{"view_html":"https://pith.science/pith/D72DBSIFWZEEFOMZE4IHZVEEBT","download_json":"https://pith.science/pith/D72DBSIFWZEEFOMZE4IHZVEEBT.json","view_paper":"https://pith.science/paper/D72DBSIF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.00734&json=true","fetch_graph":"https://pith.science/api/pith-number/D72DBSIFWZEEFOMZE4IHZVEEBT/graph.json","fetch_events":"https://pith.science/api/pith-number/D72DBSIFWZEEFOMZE4IHZVEEBT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D72DBSIFWZEEFOMZE4IHZVEEBT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D72DBSIFWZEEFOMZE4IHZVEEBT/action/storage_attestation","attest_author":"https://pith.science/pith/D72DBSIFWZEEFOMZE4IHZVEEBT/action/author_attestation","sign_citation":"https://pith.science/pith/D72DBSIFWZEEFOMZE4IHZVEEBT/action/citation_signature","submit_replication":"https://pith.science/pith/D72DBSIFWZEEFOMZE4IHZVEEBT/action/replication_record"}},"created_at":"2026-07-05T03:54:41.441379+00:00","updated_at":"2026-07-05T03:54:41.441379+00:00"}