{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:44CG5JVO5QFSQOLGZPLFJL7ZNV","short_pith_number":"pith:44CG5JVO","schema_version":"1.0","canonical_sha256":"e7046ea6aeec0b283966cbd654aff96d564e7f21e7e774ef1b08b69c0c2e5c90","source":{"kind":"arxiv","id":"2305.07583","version":3},"attestation_state":"computed","paper":{"title":"MoMo: Momentum Models for Adaptive Learning Rates","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Aaron Defazio, Fabian Schaipp, Michael Eickenberg, Robert M. Gower, Ruben Ohana","submitted_at":"2023-05-12T16:25:57Z","abstract_excerpt":"Training a modern machine learning architecture on a new task requires extensive learning-rate tuning, which comes at a high computational cost. Here we develop new Polyak-type adaptive learning rates that can be used on top of any momentum method, and require less tuning to perform well. We first develop MoMo, a Momentum Model based adaptive learning rate for SGD-M (stochastic gradient descent with momentum). MoMo uses momentum estimates of the losses and gradients sampled at each iteration to build a model of the loss function. Our model makes use of any known lower bound of the loss functio"},"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":"2305.07583","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-05-12T16:25:57Z","cross_cats_sorted":["math.OC"],"title_canon_sha256":"741bd7f615d635f7ff56a10501eb8b5f17af427a20b98ff55a8b319021b64366","abstract_canon_sha256":"6f505356e4cf80fec5a2704b7d5829918eeb232308abc35829a90ebadf560602"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:27:40.387159Z","signature_b64":"S2AXSQC8h70Q+5uyN+JZ2/ajDkWyLgKBRxSDtNY9189eUu4PM/4QgilOU0WdQG7rzJSfwwk6vCMVFFqKqvRIDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e7046ea6aeec0b283966cbd654aff96d564e7f21e7e774ef1b08b69c0c2e5c90","last_reissued_at":"2026-07-05T08:27:40.386685Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:27:40.386685Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MoMo: Momentum Models for Adaptive Learning Rates","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Aaron Defazio, Fabian Schaipp, Michael Eickenberg, Robert M. Gower, Ruben Ohana","submitted_at":"2023-05-12T16:25:57Z","abstract_excerpt":"Training a modern machine learning architecture on a new task requires extensive learning-rate tuning, which comes at a high computational cost. Here we develop new Polyak-type adaptive learning rates that can be used on top of any momentum method, and require less tuning to perform well. We first develop MoMo, a Momentum Model based adaptive learning rate for SGD-M (stochastic gradient descent with momentum). MoMo uses momentum estimates of the losses and gradients sampled at each iteration to build a model of the loss function. Our model makes use of any known lower bound of the loss functio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.07583","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/2305.07583/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":"2305.07583","created_at":"2026-07-05T08:27:40.386751+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.07583v3","created_at":"2026-07-05T08:27:40.386751+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.07583","created_at":"2026-07-05T08:27:40.386751+00:00"},{"alias_kind":"pith_short_12","alias_value":"44CG5JVO5QFS","created_at":"2026-07-05T08:27:40.386751+00:00"},{"alias_kind":"pith_short_16","alias_value":"44CG5JVO5QFSQOLG","created_at":"2026-07-05T08:27:40.386751+00:00"},{"alias_kind":"pith_short_8","alias_value":"44CG5JVO","created_at":"2026-07-05T08:27:40.386751+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.14970","citing_title":"Zero-order Parameter-free Optimization for LMO-based Methods: Novel Approach for Efficient Fine-tuning","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2510.04988","citing_title":"Adaptive Memory Momentum via a Model-Based Framework for Deep Learning Optimization","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11838","citing_title":"Gradient Clipping Beyond Vector Norms: A Spectral Approach for Matrix-Valued Parameters","ref_index":54,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/44CG5JVO5QFSQOLGZPLFJL7ZNV","json":"https://pith.science/pith/44CG5JVO5QFSQOLGZPLFJL7ZNV.json","graph_json":"https://pith.science/api/pith-number/44CG5JVO5QFSQOLGZPLFJL7ZNV/graph.json","events_json":"https://pith.science/api/pith-number/44CG5JVO5QFSQOLGZPLFJL7ZNV/events.json","paper":"https://pith.science/paper/44CG5JVO"},"agent_actions":{"view_html":"https://pith.science/pith/44CG5JVO5QFSQOLGZPLFJL7ZNV","download_json":"https://pith.science/pith/44CG5JVO5QFSQOLGZPLFJL7ZNV.json","view_paper":"https://pith.science/paper/44CG5JVO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.07583&json=true","fetch_graph":"https://pith.science/api/pith-number/44CG5JVO5QFSQOLGZPLFJL7ZNV/graph.json","fetch_events":"https://pith.science/api/pith-number/44CG5JVO5QFSQOLGZPLFJL7ZNV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/44CG5JVO5QFSQOLGZPLFJL7ZNV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/44CG5JVO5QFSQOLGZPLFJL7ZNV/action/storage_attestation","attest_author":"https://pith.science/pith/44CG5JVO5QFSQOLGZPLFJL7ZNV/action/author_attestation","sign_citation":"https://pith.science/pith/44CG5JVO5QFSQOLGZPLFJL7ZNV/action/citation_signature","submit_replication":"https://pith.science/pith/44CG5JVO5QFSQOLGZPLFJL7ZNV/action/replication_record"}},"created_at":"2026-07-05T08:27:40.386751+00:00","updated_at":"2026-07-05T08:27:40.386751+00:00"}