{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:AXCGT2YM3KGTEXOWY2E6POCKZV","short_pith_number":"pith:AXCGT2YM","schema_version":"1.0","canonical_sha256":"05c469eb0cda8d325dd6c689e7b84acd6f3f1785e59a041bcd0158bb9f2ce636","source":{"kind":"arxiv","id":"2012.10424","version":2},"attestation_state":"computed","paper":{"title":"Separation and Concentration in Deep Networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Florentin Guth, John Zarka, St\\'ephane Mallat","submitted_at":"2020-12-18T18:27:37Z","abstract_excerpt":"Numerical experiments demonstrate that deep neural network classifiers progressively separate class distributions around their mean, achieving linear separability on the training set, and increasing the Fisher discriminant ratio. We explain this mechanism with two types of operators. We prove that a rectifier without biases applied to sign-invariant tight frames can separate class means and increase Fisher ratios. On the opposite, a soft-thresholding on tight frames can reduce within-class variabilities while preserving class means. Variance reduction bounds are proved for Gaussian mixture mod"},"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":"2012.10424","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-12-18T18:27:37Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"6f0c853f9d673eb603f9e6f31070fb101ef69d1b1f84a4e318f170bf9545eddf","abstract_canon_sha256":"ffcd29129efcedae73004de2ecc8908f0b53fb02c73a569bf66ca7df005b767e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:22:57.755256Z","signature_b64":"wmmYVjsUnqp0H6q0tcWusp0SfMzqoXD7un4l6MOUHWzRhgwYgOkZ6H9m6fbsOqM2TSgJ1OUhPiQ/EXomHqlFBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"05c469eb0cda8d325dd6c689e7b84acd6f3f1785e59a041bcd0158bb9f2ce636","last_reissued_at":"2026-07-05T02:22:57.754779Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:22:57.754779Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Separation and Concentration in Deep Networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Florentin Guth, John Zarka, St\\'ephane Mallat","submitted_at":"2020-12-18T18:27:37Z","abstract_excerpt":"Numerical experiments demonstrate that deep neural network classifiers progressively separate class distributions around their mean, achieving linear separability on the training set, and increasing the Fisher discriminant ratio. We explain this mechanism with two types of operators. We prove that a rectifier without biases applied to sign-invariant tight frames can separate class means and increase Fisher ratios. On the opposite, a soft-thresholding on tight frames can reduce within-class variabilities while preserving class means. Variance reduction bounds are proved for Gaussian mixture mod"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2012.10424","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/2012.10424/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":"2012.10424","created_at":"2026-07-05T02:22:57.754839+00:00"},{"alias_kind":"arxiv_version","alias_value":"2012.10424v2","created_at":"2026-07-05T02:22:57.754839+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2012.10424","created_at":"2026-07-05T02:22:57.754839+00:00"},{"alias_kind":"pith_short_12","alias_value":"AXCGT2YM3KGT","created_at":"2026-07-05T02:22:57.754839+00:00"},{"alias_kind":"pith_short_16","alias_value":"AXCGT2YM3KGTEXOW","created_at":"2026-07-05T02:22:57.754839+00:00"},{"alias_kind":"pith_short_8","alias_value":"AXCGT2YM","created_at":"2026-07-05T02:22:57.754839+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.30813","citing_title":"Gradient Smoothing: Coupling Layer-wise Updates for Improved Optimization","ref_index":40,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AXCGT2YM3KGTEXOWY2E6POCKZV","json":"https://pith.science/pith/AXCGT2YM3KGTEXOWY2E6POCKZV.json","graph_json":"https://pith.science/api/pith-number/AXCGT2YM3KGTEXOWY2E6POCKZV/graph.json","events_json":"https://pith.science/api/pith-number/AXCGT2YM3KGTEXOWY2E6POCKZV/events.json","paper":"https://pith.science/paper/AXCGT2YM"},"agent_actions":{"view_html":"https://pith.science/pith/AXCGT2YM3KGTEXOWY2E6POCKZV","download_json":"https://pith.science/pith/AXCGT2YM3KGTEXOWY2E6POCKZV.json","view_paper":"https://pith.science/paper/AXCGT2YM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2012.10424&json=true","fetch_graph":"https://pith.science/api/pith-number/AXCGT2YM3KGTEXOWY2E6POCKZV/graph.json","fetch_events":"https://pith.science/api/pith-number/AXCGT2YM3KGTEXOWY2E6POCKZV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AXCGT2YM3KGTEXOWY2E6POCKZV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AXCGT2YM3KGTEXOWY2E6POCKZV/action/storage_attestation","attest_author":"https://pith.science/pith/AXCGT2YM3KGTEXOWY2E6POCKZV/action/author_attestation","sign_citation":"https://pith.science/pith/AXCGT2YM3KGTEXOWY2E6POCKZV/action/citation_signature","submit_replication":"https://pith.science/pith/AXCGT2YM3KGTEXOWY2E6POCKZV/action/replication_record"}},"created_at":"2026-07-05T02:22:57.754839+00:00","updated_at":"2026-07-05T02:22:57.754839+00:00"}