{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:RIH7YYHYYZKYI3DGQWJ6HCGVVB","short_pith_number":"pith:RIH7YYHY","schema_version":"1.0","canonical_sha256":"8a0ffc60f8c655846c668593e388d5a86024e29ef5e8e7a7f72b0d4145d0c912","source":{"kind":"arxiv","id":"2001.02992","version":1},"attestation_state":"computed","paper":{"title":"Poly-time universality and limitations of deep learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CC","cs.IT","math.IT","stat.ML"],"primary_cat":"cs.LG","authors_text":"Colin Sandon, Emmanuel Abbe","submitted_at":"2020-01-07T08:31:50Z","abstract_excerpt":"The goal of this paper is to characterize function distributions that deep learning can or cannot learn in poly-time. A universality result is proved for SGD-based deep learning and a non-universality result is proved for GD-based deep learning; this also gives a separation between SGD-based deep learning and statistical query algorithms:\n  (1) {\\it Deep learning with SGD is efficiently universal.} Any function distribution that can be learned from samples in poly-time can also be learned by a poly-size neural net trained with SGD on a poly-time initialization with poly-steps, poly-rate and po"},"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":"2001.02992","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-01-07T08:31:50Z","cross_cats_sorted":["cs.CC","cs.IT","math.IT","stat.ML"],"title_canon_sha256":"6289cb90f8317abbbcc07663b4dbf1ec564cc5d9476754adad6dc433e41757e6","abstract_canon_sha256":"5c93c13b5f0643d050e3c44e3d285933af4c6c0548e5a5fbd93d1ba6eb0fee46"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T00:32:32.516765Z","signature_b64":"BjNzQHj/kfP8GLduClJMtiuoHQz8hkXapc224qj9Zau3S2i2gvU3H2giW7xHq551rlipZ+KfswqsYjBUcRplBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8a0ffc60f8c655846c668593e388d5a86024e29ef5e8e7a7f72b0d4145d0c912","last_reissued_at":"2026-07-05T00:32:32.516309Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T00:32:32.516309Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Poly-time universality and limitations of deep learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CC","cs.IT","math.IT","stat.ML"],"primary_cat":"cs.LG","authors_text":"Colin Sandon, Emmanuel Abbe","submitted_at":"2020-01-07T08:31:50Z","abstract_excerpt":"The goal of this paper is to characterize function distributions that deep learning can or cannot learn in poly-time. A universality result is proved for SGD-based deep learning and a non-universality result is proved for GD-based deep learning; this also gives a separation between SGD-based deep learning and statistical query algorithms:\n  (1) {\\it Deep learning with SGD is efficiently universal.} Any function distribution that can be learned from samples in poly-time can also be learned by a poly-size neural net trained with SGD on a poly-time initialization with poly-steps, poly-rate and po"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2001.02992","kind":"arxiv","version":1},"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/2001.02992/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":"2001.02992","created_at":"2026-07-05T00:32:32.516379+00:00"},{"alias_kind":"arxiv_version","alias_value":"2001.02992v1","created_at":"2026-07-05T00:32:32.516379+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2001.02992","created_at":"2026-07-05T00:32:32.516379+00:00"},{"alias_kind":"pith_short_12","alias_value":"RIH7YYHYYZKY","created_at":"2026-07-05T00:32:32.516379+00:00"},{"alias_kind":"pith_short_16","alias_value":"RIH7YYHYYZKYI3DG","created_at":"2026-07-05T00:32:32.516379+00:00"},{"alias_kind":"pith_short_8","alias_value":"RIH7YYHY","created_at":"2026-07-05T00:32:32.516379+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2510.26745","citing_title":"Deep sequence models tend to memorize geometrically; it is unclear why","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RIH7YYHYYZKYI3DGQWJ6HCGVVB","json":"https://pith.science/pith/RIH7YYHYYZKYI3DGQWJ6HCGVVB.json","graph_json":"https://pith.science/api/pith-number/RIH7YYHYYZKYI3DGQWJ6HCGVVB/graph.json","events_json":"https://pith.science/api/pith-number/RIH7YYHYYZKYI3DGQWJ6HCGVVB/events.json","paper":"https://pith.science/paper/RIH7YYHY"},"agent_actions":{"view_html":"https://pith.science/pith/RIH7YYHYYZKYI3DGQWJ6HCGVVB","download_json":"https://pith.science/pith/RIH7YYHYYZKYI3DGQWJ6HCGVVB.json","view_paper":"https://pith.science/paper/RIH7YYHY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2001.02992&json=true","fetch_graph":"https://pith.science/api/pith-number/RIH7YYHYYZKYI3DGQWJ6HCGVVB/graph.json","fetch_events":"https://pith.science/api/pith-number/RIH7YYHYYZKYI3DGQWJ6HCGVVB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RIH7YYHYYZKYI3DGQWJ6HCGVVB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RIH7YYHYYZKYI3DGQWJ6HCGVVB/action/storage_attestation","attest_author":"https://pith.science/pith/RIH7YYHYYZKYI3DGQWJ6HCGVVB/action/author_attestation","sign_citation":"https://pith.science/pith/RIH7YYHYYZKYI3DGQWJ6HCGVVB/action/citation_signature","submit_replication":"https://pith.science/pith/RIH7YYHYYZKYI3DGQWJ6HCGVVB/action/replication_record"}},"created_at":"2026-07-05T00:32:32.516379+00:00","updated_at":"2026-07-05T00:32:32.516379+00:00"}