{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:VT47ORQQZ5XDGDH7W77RPP4BDB","short_pith_number":"pith:VT47ORQQ","schema_version":"1.0","canonical_sha256":"acf9f74610cf6e330cffb7ff17bf81185dbb9cf98f4261a7997d61684d64a411","source":{"kind":"arxiv","id":"1808.04295","version":4},"attestation_state":"computed","paper":{"title":"Understanding training and generalization in deep learning by Fourier analysis","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","math.OC","math.ST","stat.ML","stat.TH"],"primary_cat":"cs.LG","authors_text":"Zhiqin John Xu","submitted_at":"2018-08-13T15:40:41Z","abstract_excerpt":"Background: It is still an open research area to theoretically understand why Deep Neural Networks (DNNs)---equipped with many more parameters than training data and trained by (stochastic) gradient-based methods---often achieve remarkably low generalization error. Contribution: We study DNN training by Fourier analysis. Our theoretical framework explains: i) DNN with (stochastic) gradient-based methods often endows low-frequency components of the target function with a higher priority during the training; ii) Small initialization leads to good generalization ability of DNN while preserving th"},"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":"1808.04295","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2018-08-13T15:40:41Z","cross_cats_sorted":["cs.AI","math.OC","math.ST","stat.ML","stat.TH"],"title_canon_sha256":"eb0a34147a46c5e44668ad3de40bfaac0a0673d4025d1bf03c5ea005058e8c64","abstract_canon_sha256":"ade06a957ac648d4c55870a59bbeb8b157c4f189806b0990cb639ff2b13ad872"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:59:38.457248Z","signature_b64":"hPAnvtiY0EwiRwmRwz1gEwB0gG96XVde0oiXpRvTnJlz91USGOFFi/ZcJAm0psNlc0dNfwj7v2QaIlVNqCXRAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"acf9f74610cf6e330cffb7ff17bf81185dbb9cf98f4261a7997d61684d64a411","last_reissued_at":"2026-05-17T23:59:38.456614Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:59:38.456614Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Understanding training and generalization in deep learning by Fourier analysis","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","math.OC","math.ST","stat.ML","stat.TH"],"primary_cat":"cs.LG","authors_text":"Zhiqin John Xu","submitted_at":"2018-08-13T15:40:41Z","abstract_excerpt":"Background: It is still an open research area to theoretically understand why Deep Neural Networks (DNNs)---equipped with many more parameters than training data and trained by (stochastic) gradient-based methods---often achieve remarkably low generalization error. Contribution: We study DNN training by Fourier analysis. Our theoretical framework explains: i) DNN with (stochastic) gradient-based methods often endows low-frequency components of the target function with a higher priority during the training; ii) Small initialization leads to good generalization ability of DNN while preserving th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1808.04295","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":""},"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":"1808.04295","created_at":"2026-05-17T23:59:38.456721+00:00"},{"alias_kind":"arxiv_version","alias_value":"1808.04295v4","created_at":"2026-05-17T23:59:38.456721+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1808.04295","created_at":"2026-05-17T23:59:38.456721+00:00"},{"alias_kind":"pith_short_12","alias_value":"VT47ORQQZ5XD","created_at":"2026-05-18T12:32:59.047623+00:00"},{"alias_kind":"pith_short_16","alias_value":"VT47ORQQZ5XDGDH7","created_at":"2026-05-18T12:32:59.047623+00:00"},{"alias_kind":"pith_short_8","alias_value":"VT47ORQQ","created_at":"2026-05-18T12:32:59.047623+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"1906.09235","citing_title":"Theory of the Frequency Principle for General Deep Neural Networks","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2510.26745","citing_title":"Deep sequence models tend to memorize geometrically; it is unclear why","ref_index":190,"is_internal_anchor":true},{"citing_arxiv_id":"2508.15198","citing_title":"Frequency-adaptive tensor neural networks for high-dimensional multi-scale problems","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12709","citing_title":"Spectral Energy Centroid: a Metric for Improving Performance and Analyzing Spectral Bias in Implicit Neural Representations","ref_index":22,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VT47ORQQZ5XDGDH7W77RPP4BDB","json":"https://pith.science/pith/VT47ORQQZ5XDGDH7W77RPP4BDB.json","graph_json":"https://pith.science/api/pith-number/VT47ORQQZ5XDGDH7W77RPP4BDB/graph.json","events_json":"https://pith.science/api/pith-number/VT47ORQQZ5XDGDH7W77RPP4BDB/events.json","paper":"https://pith.science/paper/VT47ORQQ"},"agent_actions":{"view_html":"https://pith.science/pith/VT47ORQQZ5XDGDH7W77RPP4BDB","download_json":"https://pith.science/pith/VT47ORQQZ5XDGDH7W77RPP4BDB.json","view_paper":"https://pith.science/paper/VT47ORQQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1808.04295&json=true","fetch_graph":"https://pith.science/api/pith-number/VT47ORQQZ5XDGDH7W77RPP4BDB/graph.json","fetch_events":"https://pith.science/api/pith-number/VT47ORQQZ5XDGDH7W77RPP4BDB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VT47ORQQZ5XDGDH7W77RPP4BDB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VT47ORQQZ5XDGDH7W77RPP4BDB/action/storage_attestation","attest_author":"https://pith.science/pith/VT47ORQQZ5XDGDH7W77RPP4BDB/action/author_attestation","sign_citation":"https://pith.science/pith/VT47ORQQZ5XDGDH7W77RPP4BDB/action/citation_signature","submit_replication":"https://pith.science/pith/VT47ORQQZ5XDGDH7W77RPP4BDB/action/replication_record"}},"created_at":"2026-05-17T23:59:38.456721+00:00","updated_at":"2026-05-17T23:59:38.456721+00:00"}