{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:OUP3553CJMGMOQNID5VOU37KDZ","short_pith_number":"pith:OUP3553C","schema_version":"1.0","canonical_sha256":"751fbef7624b0cc741a81f6aea6fea1e4aad74f6232e5190c2e0bef91d1eaa21","source":{"kind":"arxiv","id":"2105.12806","version":4},"attestation_state":"computed","paper":{"title":"A Universal Law of Robustness via Isoperimetry","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Mark Sellke, S\\'ebastien Bubeck","submitted_at":"2021-05-26T19:49:47Z","abstract_excerpt":"Classically, data interpolation with a parametrized model class is possible as long as the number of parameters is larger than the number of equations to be satisfied. A puzzling phenomenon in deep learning is that models are trained with many more parameters than what this classical theory would suggest. We propose a partial theoretical explanation for this phenomenon. We prove that for a broad class of data distributions and model classes, overparametrization is necessary if one wants to interpolate the data smoothly. Namely we show that smooth interpolation requires $d$ times more parameter"},"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":"2105.12806","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-05-26T19:49:47Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"fe7c84a965b15bb4cd9690db08b716778cdde10a78543ac8f6ae7372753df551","abstract_canon_sha256":"7e5e311185dc631c2d0320ccf194243c2ce2a07b5b524923baabf672a64b972e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:28:04.833013Z","signature_b64":"UJhfzE4fAiDLLl6N7OJw0d3oNO3LT22IylR/Kl0g4sQKZojRu3EyTlhAGTILX+EG1mHx2xYT9/lDM6Wq9rquBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"751fbef7624b0cc741a81f6aea6fea1e4aad74f6232e5190c2e0bef91d1eaa21","last_reissued_at":"2026-07-05T05:28:04.832585Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:28:04.832585Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Universal Law of Robustness via Isoperimetry","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Mark Sellke, S\\'ebastien Bubeck","submitted_at":"2021-05-26T19:49:47Z","abstract_excerpt":"Classically, data interpolation with a parametrized model class is possible as long as the number of parameters is larger than the number of equations to be satisfied. A puzzling phenomenon in deep learning is that models are trained with many more parameters than what this classical theory would suggest. We propose a partial theoretical explanation for this phenomenon. We prove that for a broad class of data distributions and model classes, overparametrization is necessary if one wants to interpolate the data smoothly. Namely we show that smooth interpolation requires $d$ times more parameter"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2105.12806","kind":"arxiv","version":4},"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/2105.12806/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":"2105.12806","created_at":"2026-07-05T05:28:04.832641+00:00"},{"alias_kind":"arxiv_version","alias_value":"2105.12806v4","created_at":"2026-07-05T05:28:04.832641+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2105.12806","created_at":"2026-07-05T05:28:04.832641+00:00"},{"alias_kind":"pith_short_12","alias_value":"OUP3553CJMGM","created_at":"2026-07-05T05:28:04.832641+00:00"},{"alias_kind":"pith_short_16","alias_value":"OUP3553CJMGMOQNI","created_at":"2026-07-05T05:28:04.832641+00:00"},{"alias_kind":"pith_short_8","alias_value":"OUP3553C","created_at":"2026-07-05T05:28:04.832641+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07778","citing_title":"A law of robustness for two-layer neural networks with arbitrary weights","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20866","citing_title":"LOSCAR-SGD: Local SGD with Communication-Computation Overlap and Delay-Corrected Sparse Model Averaging","ref_index":140,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18174","citing_title":"Ringmaster LMO: Asynchronous Linear Minimization Oracle Momentum Method","ref_index":138,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13434","citing_title":"Rescaled Asynchronous SGD: Optimal Distributed Optimization under Data and System Heterogeneity","ref_index":247,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08871","citing_title":"Rennala MVR: Improved Time Complexity for Parallel Stochastic Optimization via Momentum-Based Variance Reduction","ref_index":136,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07795","citing_title":"Scalable Distributed Stochastic Optimization via Bidirectional Compression: Beyond Pessimistic Limits","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03045","citing_title":"TCD-Arena: Assessing Robustness of Time Series Causal Discovery Methods Against Assumption Violations","ref_index":255,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OUP3553CJMGMOQNID5VOU37KDZ","json":"https://pith.science/pith/OUP3553CJMGMOQNID5VOU37KDZ.json","graph_json":"https://pith.science/api/pith-number/OUP3553CJMGMOQNID5VOU37KDZ/graph.json","events_json":"https://pith.science/api/pith-number/OUP3553CJMGMOQNID5VOU37KDZ/events.json","paper":"https://pith.science/paper/OUP3553C"},"agent_actions":{"view_html":"https://pith.science/pith/OUP3553CJMGMOQNID5VOU37KDZ","download_json":"https://pith.science/pith/OUP3553CJMGMOQNID5VOU37KDZ.json","view_paper":"https://pith.science/paper/OUP3553C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2105.12806&json=true","fetch_graph":"https://pith.science/api/pith-number/OUP3553CJMGMOQNID5VOU37KDZ/graph.json","fetch_events":"https://pith.science/api/pith-number/OUP3553CJMGMOQNID5VOU37KDZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OUP3553CJMGMOQNID5VOU37KDZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OUP3553CJMGMOQNID5VOU37KDZ/action/storage_attestation","attest_author":"https://pith.science/pith/OUP3553CJMGMOQNID5VOU37KDZ/action/author_attestation","sign_citation":"https://pith.science/pith/OUP3553CJMGMOQNID5VOU37KDZ/action/citation_signature","submit_replication":"https://pith.science/pith/OUP3553CJMGMOQNID5VOU37KDZ/action/replication_record"}},"created_at":"2026-07-05T05:28:04.832641+00:00","updated_at":"2026-07-05T05:28:04.832641+00:00"}