{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2017:TTSE4RZ3V7QL73ZY6HKYBRJ5XT","short_pith_number":"pith:TTSE4RZ3","schema_version":"1.0","canonical_sha256":"9ce44e473bafe0bfef38f1d580c53dbce0f55f10766611365c77a06b2df85446","source":{"kind":"arxiv","id":"1708.02691","version":3},"attestation_state":"computed","paper":{"title":"Universal Function Approximation by Deep Neural Nets with Bounded Width and ReLU Activations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CG","cs.LG","math.FA","math.ST","stat.TH"],"primary_cat":"stat.ML","authors_text":"Boris Hanin","submitted_at":"2017-08-09T01:37:21Z","abstract_excerpt":"This article concerns the expressive power of depth in neural nets with ReLU activations and bounded width. We are particularly interested in the following questions: what is the minimal width $w_{\\text{min}}(d)$ so that ReLU nets of width $w_{\\text{min}}(d)$ (and arbitrary depth) can approximate any continuous function on the unit cube $[0,1]^d$ aribitrarily well? For ReLU nets near this minimal width, what can one say about the depth necessary to approximate a given function? Our approach to this paper is based on the observation that, due to the convexity of the ReLU activation, ReLU nets 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":"1708.02691","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2017-08-09T01:37:21Z","cross_cats_sorted":["cs.CG","cs.LG","math.FA","math.ST","stat.TH"],"title_canon_sha256":"ac093a691aa3d220ef67b248554e694f141eeeb158092dd749278c5381e9adea","abstract_canon_sha256":"1b41af95f97357f6a8e2817e51063d45a2df781560aca167c0cb362033d4655e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T00:13:11.754765Z","signature_b64":"uCQP7lXyoTlg1S9wDxUZXZGFYevz3m+7uvCa4N9recv4mEcpSE994TZLFmfgBdA393wLBYgjZTgGdqpyJJaUBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9ce44e473bafe0bfef38f1d580c53dbce0f55f10766611365c77a06b2df85446","last_reissued_at":"2026-07-05T00:13:11.754410Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T00:13:11.754410Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Universal Function Approximation by Deep Neural Nets with Bounded Width and ReLU Activations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CG","cs.LG","math.FA","math.ST","stat.TH"],"primary_cat":"stat.ML","authors_text":"Boris Hanin","submitted_at":"2017-08-09T01:37:21Z","abstract_excerpt":"This article concerns the expressive power of depth in neural nets with ReLU activations and bounded width. We are particularly interested in the following questions: what is the minimal width $w_{\\text{min}}(d)$ so that ReLU nets of width $w_{\\text{min}}(d)$ (and arbitrary depth) can approximate any continuous function on the unit cube $[0,1]^d$ aribitrarily well? For ReLU nets near this minimal width, what can one say about the depth necessary to approximate a given function? Our approach to this paper is based on the observation that, due to the convexity of the ReLU activation, ReLU nets a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1708.02691","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/1708.02691/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":"1708.02691","created_at":"2026-07-05T00:13:11.754466+00:00"},{"alias_kind":"arxiv_version","alias_value":"1708.02691v3","created_at":"2026-07-05T00:13:11.754466+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1708.02691","created_at":"2026-07-05T00:13:11.754466+00:00"},{"alias_kind":"pith_short_12","alias_value":"TTSE4RZ3V7QL","created_at":"2026-07-05T00:13:11.754466+00:00"},{"alias_kind":"pith_short_16","alias_value":"TTSE4RZ3V7QL73ZY","created_at":"2026-07-05T00:13:11.754466+00:00"},{"alias_kind":"pith_short_8","alias_value":"TTSE4RZ3","created_at":"2026-07-05T00:13:11.754466+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.10089","citing_title":"A Theory on Flow Matching with Neural Networks","ref_index":211,"is_internal_anchor":false},{"citing_arxiv_id":"1906.11235","citing_title":"Invariance-inducing regularization using worst-case transformations suffices to boost accuracy and spatial robustness","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2112.14523","citing_title":"Deep neural network approximation theory for high-dimensional functions","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"1910.03193","citing_title":"DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operators","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2401.01335","citing_title":"Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models","ref_index":185,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TTSE4RZ3V7QL73ZY6HKYBRJ5XT","json":"https://pith.science/pith/TTSE4RZ3V7QL73ZY6HKYBRJ5XT.json","graph_json":"https://pith.science/api/pith-number/TTSE4RZ3V7QL73ZY6HKYBRJ5XT/graph.json","events_json":"https://pith.science/api/pith-number/TTSE4RZ3V7QL73ZY6HKYBRJ5XT/events.json","paper":"https://pith.science/paper/TTSE4RZ3"},"agent_actions":{"view_html":"https://pith.science/pith/TTSE4RZ3V7QL73ZY6HKYBRJ5XT","download_json":"https://pith.science/pith/TTSE4RZ3V7QL73ZY6HKYBRJ5XT.json","view_paper":"https://pith.science/paper/TTSE4RZ3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1708.02691&json=true","fetch_graph":"https://pith.science/api/pith-number/TTSE4RZ3V7QL73ZY6HKYBRJ5XT/graph.json","fetch_events":"https://pith.science/api/pith-number/TTSE4RZ3V7QL73ZY6HKYBRJ5XT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TTSE4RZ3V7QL73ZY6HKYBRJ5XT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TTSE4RZ3V7QL73ZY6HKYBRJ5XT/action/storage_attestation","attest_author":"https://pith.science/pith/TTSE4RZ3V7QL73ZY6HKYBRJ5XT/action/author_attestation","sign_citation":"https://pith.science/pith/TTSE4RZ3V7QL73ZY6HKYBRJ5XT/action/citation_signature","submit_replication":"https://pith.science/pith/TTSE4RZ3V7QL73ZY6HKYBRJ5XT/action/replication_record"}},"created_at":"2026-07-05T00:13:11.754466+00:00","updated_at":"2026-07-05T00:13:11.754466+00:00"}