{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:ET5PZQGGNSD3CNHDEYWRHD4VLP","short_pith_number":"pith:ET5PZQGG","schema_version":"1.0","canonical_sha256":"24fafcc0c66c87b134e3262d138f955bd749e863dd5b4cabf4076a858c5ac45e","source":{"kind":"arxiv","id":"2102.07870","version":3},"attestation_state":"computed","paper":{"title":"Momentum Residual Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Gabriel Peyr\\'e, Mathieu Blondel, Michael E. Sander, Pierre Ablin","submitted_at":"2021-02-15T22:24:52Z","abstract_excerpt":"The training of deep residual neural networks (ResNets) with backpropagation has a memory cost that increases linearly with respect to the depth of the network. A way to circumvent this issue is to use reversible architectures. In this paper, we propose to change the forward rule of a ResNet by adding a momentum term. The resulting networks, momentum residual neural networks (Momentum ResNets), are invertible. Unlike previous invertible architectures, they can be used as a drop-in replacement for any existing ResNet block. We show that Momentum ResNets can be interpreted in the infinitesimal s"},"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":"2102.07870","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2021-02-15T22:24:52Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"03109ac6c4cfb827dfce287d1edac857e4da166fdaccdb8b2640476c3a3a0e52","abstract_canon_sha256":"a560c964094c6799a4b39f7f84bc599964bb308e4e06ddeaea4dada7dc69ffbb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:59:55.329160Z","signature_b64":"V38P2PBUJGrVitKAQBFMSEuORiwYC/fqBOkw6lTQGVVSrytsaE7uWWoiIj+0EH3Ecz90Ri5fa44spBnTjM3/Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"24fafcc0c66c87b134e3262d138f955bd749e863dd5b4cabf4076a858c5ac45e","last_reissued_at":"2026-07-05T02:59:55.328707Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:59:55.328707Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Momentum Residual Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Gabriel Peyr\\'e, Mathieu Blondel, Michael E. Sander, Pierre Ablin","submitted_at":"2021-02-15T22:24:52Z","abstract_excerpt":"The training of deep residual neural networks (ResNets) with backpropagation has a memory cost that increases linearly with respect to the depth of the network. A way to circumvent this issue is to use reversible architectures. In this paper, we propose to change the forward rule of a ResNet by adding a momentum term. The resulting networks, momentum residual neural networks (Momentum ResNets), are invertible. Unlike previous invertible architectures, they can be used as a drop-in replacement for any existing ResNet block. We show that Momentum ResNets can be interpreted in the infinitesimal s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2102.07870","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/2102.07870/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":"2102.07870","created_at":"2026-07-05T02:59:55.328762+00:00"},{"alias_kind":"arxiv_version","alias_value":"2102.07870v3","created_at":"2026-07-05T02:59:55.328762+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2102.07870","created_at":"2026-07-05T02:59:55.328762+00:00"},{"alias_kind":"pith_short_12","alias_value":"ET5PZQGGNSD3","created_at":"2026-07-05T02:59:55.328762+00:00"},{"alias_kind":"pith_short_16","alias_value":"ET5PZQGGNSD3CNHD","created_at":"2026-07-05T02:59:55.328762+00:00"},{"alias_kind":"pith_short_8","alias_value":"ET5PZQGG","created_at":"2026-07-05T02:59:55.328762+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.07244","citing_title":"The Influence of the Memory Capacity of Neural DDEs on the Universal Approximation Property","ref_index":47,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ET5PZQGGNSD3CNHDEYWRHD4VLP","json":"https://pith.science/pith/ET5PZQGGNSD3CNHDEYWRHD4VLP.json","graph_json":"https://pith.science/api/pith-number/ET5PZQGGNSD3CNHDEYWRHD4VLP/graph.json","events_json":"https://pith.science/api/pith-number/ET5PZQGGNSD3CNHDEYWRHD4VLP/events.json","paper":"https://pith.science/paper/ET5PZQGG"},"agent_actions":{"view_html":"https://pith.science/pith/ET5PZQGGNSD3CNHDEYWRHD4VLP","download_json":"https://pith.science/pith/ET5PZQGGNSD3CNHDEYWRHD4VLP.json","view_paper":"https://pith.science/paper/ET5PZQGG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2102.07870&json=true","fetch_graph":"https://pith.science/api/pith-number/ET5PZQGGNSD3CNHDEYWRHD4VLP/graph.json","fetch_events":"https://pith.science/api/pith-number/ET5PZQGGNSD3CNHDEYWRHD4VLP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ET5PZQGGNSD3CNHDEYWRHD4VLP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ET5PZQGGNSD3CNHDEYWRHD4VLP/action/storage_attestation","attest_author":"https://pith.science/pith/ET5PZQGGNSD3CNHDEYWRHD4VLP/action/author_attestation","sign_citation":"https://pith.science/pith/ET5PZQGGNSD3CNHDEYWRHD4VLP/action/citation_signature","submit_replication":"https://pith.science/pith/ET5PZQGGNSD3CNHDEYWRHD4VLP/action/replication_record"}},"created_at":"2026-07-05T02:59:55.328762+00:00","updated_at":"2026-07-05T02:59:55.328762+00:00"}