{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:PAF737H7IKZFRATKQMYWOSIWPS","short_pith_number":"pith:PAF737H7","schema_version":"1.0","canonical_sha256":"780bfdfcff42b258826a83316749167cbd19e8dcaeff0283c883234ad422cf95","source":{"kind":"arxiv","id":"2402.01868","version":2},"attestation_state":"computed","paper":{"title":"Challenges in Training PINNs: A Loss Landscape Perspective","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Lu Lu, Madeleine Udell, Pratik Rathore, Weimu Lei, Zachary Frangella","submitted_at":"2024-02-02T19:46:43Z","abstract_excerpt":"This paper explores challenges in training Physics-Informed Neural Networks (PINNs), emphasizing the role of the loss landscape in the training process. We examine difficulties in minimizing the PINN loss function, particularly due to ill-conditioning caused by differential operators in the residual term. We compare gradient-based optimizers Adam, L-BFGS, and their combination Adam+L-BFGS, showing the superiority of Adam+L-BFGS, and introduce a novel second-order optimizer, NysNewton-CG (NNCG), which significantly improves PINN performance. Theoretically, our work elucidates the connection bet"},"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":"2402.01868","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-02T19:46:43Z","cross_cats_sorted":["math.OC","stat.ML"],"title_canon_sha256":"7e03e4925d198a2e0489a979f7f1cca9530f5b45659f6a46192128b8813ed859","abstract_canon_sha256":"fca7ba06b3ae55de6efa51b7251c64886ac110fda23029c06b865881b3a6c9c7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:26:43.622210Z","signature_b64":"gqKPdy0Pfeq/wniyWLOK2hol5k5ILyCLdCC4LXWUR2i7xQGwYdZb2a1MVv8I+b/kzJ17Uv4iOSnncknoqCljAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"780bfdfcff42b258826a83316749167cbd19e8dcaeff0283c883234ad422cf95","last_reissued_at":"2026-07-05T08:26:43.621720Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:26:43.621720Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Challenges in Training PINNs: A Loss Landscape Perspective","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Lu Lu, Madeleine Udell, Pratik Rathore, Weimu Lei, Zachary Frangella","submitted_at":"2024-02-02T19:46:43Z","abstract_excerpt":"This paper explores challenges in training Physics-Informed Neural Networks (PINNs), emphasizing the role of the loss landscape in the training process. We examine difficulties in minimizing the PINN loss function, particularly due to ill-conditioning caused by differential operators in the residual term. We compare gradient-based optimizers Adam, L-BFGS, and their combination Adam+L-BFGS, showing the superiority of Adam+L-BFGS, and introduce a novel second-order optimizer, NysNewton-CG (NNCG), which significantly improves PINN performance. Theoretically, our work elucidates the connection bet"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.01868","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/2402.01868/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":"2402.01868","created_at":"2026-07-05T08:26:43.621779+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.01868v2","created_at":"2026-07-05T08:26:43.621779+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.01868","created_at":"2026-07-05T08:26:43.621779+00:00"},{"alias_kind":"pith_short_12","alias_value":"PAF737H7IKZF","created_at":"2026-07-05T08:26:43.621779+00:00"},{"alias_kind":"pith_short_16","alias_value":"PAF737H7IKZFRATK","created_at":"2026-07-05T08:26:43.621779+00:00"},{"alias_kind":"pith_short_8","alias_value":"PAF737H7","created_at":"2026-07-05T08:26:43.621779+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":13,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.05940","citing_title":"Seed-Robust PINN Determination of $s$-Wave Bound States and Jost-Function-Based vertex constants in $_{\\Lambda}^{208}$Pb","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06171","citing_title":"Effective Dimensionality as an Operator Invariant for Physics-Preserving Constraint Adaptation in Physics-Informed Neural Networks","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03210","citing_title":"Critical evaluation of PINN for FWD inverse analysis and differentiable FEM as an alternative","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2503.23729","citing_title":"Integral regularization PINNs for evolution equations","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15254","citing_title":"Curriculum Learning of Physics-Informed Neural Networks based on Spatial Correlation","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2602.09317","citing_title":"SnareNet: Flexible Repair Layers for Neural Networks with Hard Constraints","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22862","citing_title":"Physics-Informed Neural Networks for Solving Two-Flavor Neutrino Oscillations in Vacuum and Matter Environments for Atmospheric and Reactor Neutrinos","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08408","citing_title":"AdamFLIP: Adaptive Momentum Feedback Linearization Optimization for Hard Constrained PINN Training","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23625","citing_title":"Physics informed operator learning of parameter dependent spectra","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23528","citing_title":"When PINNs Go Wrong: Pseudo-Time Stepping Against Spurious Solutions","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22862","citing_title":"Physics-Informed Neural Networks for Solving Two-Flavor Neutrino Oscillations in Vacuum and Matter Environments for Atmospheric and Reactor Neutrinos","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08869","citing_title":"Adaptive Randomized Neural Networks with Locally Activation Function: Theory and Algorithm for Solving PDEs","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14562","citing_title":"Material-agnostic temperature field prediction for metal additive manufacturing via a parametric PINN framework","ref_index":48,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PAF737H7IKZFRATKQMYWOSIWPS","json":"https://pith.science/pith/PAF737H7IKZFRATKQMYWOSIWPS.json","graph_json":"https://pith.science/api/pith-number/PAF737H7IKZFRATKQMYWOSIWPS/graph.json","events_json":"https://pith.science/api/pith-number/PAF737H7IKZFRATKQMYWOSIWPS/events.json","paper":"https://pith.science/paper/PAF737H7"},"agent_actions":{"view_html":"https://pith.science/pith/PAF737H7IKZFRATKQMYWOSIWPS","download_json":"https://pith.science/pith/PAF737H7IKZFRATKQMYWOSIWPS.json","view_paper":"https://pith.science/paper/PAF737H7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.01868&json=true","fetch_graph":"https://pith.science/api/pith-number/PAF737H7IKZFRATKQMYWOSIWPS/graph.json","fetch_events":"https://pith.science/api/pith-number/PAF737H7IKZFRATKQMYWOSIWPS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PAF737H7IKZFRATKQMYWOSIWPS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PAF737H7IKZFRATKQMYWOSIWPS/action/storage_attestation","attest_author":"https://pith.science/pith/PAF737H7IKZFRATKQMYWOSIWPS/action/author_attestation","sign_citation":"https://pith.science/pith/PAF737H7IKZFRATKQMYWOSIWPS/action/citation_signature","submit_replication":"https://pith.science/pith/PAF737H7IKZFRATKQMYWOSIWPS/action/replication_record"}},"created_at":"2026-07-05T08:26:43.621779+00:00","updated_at":"2026-07-05T08:26:43.621779+00:00"}