{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:3R44QGERLU3MWS5MCUBKUXTJLC","short_pith_number":"pith:3R44QGER","schema_version":"1.0","canonical_sha256":"dc79c818915d36cb4bac1502aa5e6958bda07b39dd3931c4b92e7ad1655da9ae","source":{"kind":"arxiv","id":"2003.04286","version":2},"attestation_state":"computed","paper":{"title":"Manifold Regularization for Locally Stable Deep Neural Networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.LG","cs.NE"],"primary_cat":"stat.ML","authors_text":"Charles Jin, Martin Rinard","submitted_at":"2020-03-09T17:45:44Z","abstract_excerpt":"We apply concepts from manifold regularization to develop new regularization techniques for training locally stable deep neural networks. Our regularizers are based on a sparsification of the graph Laplacian which holds with high probability when the data is sparse in high dimensions, as is common in deep learning. Empirically, our networks exhibit stability in a diverse set of perturbation models, including $\\ell_2$, $\\ell_\\infty$, and Wasserstein-based perturbations; in particular, we achieve 40% adversarial accuracy on CIFAR-10 against an adaptive PGD attack using $\\ell_\\infty$ perturbation"},"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":"2003.04286","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2020-03-09T17:45:44Z","cross_cats_sorted":["cs.CV","cs.LG","cs.NE"],"title_canon_sha256":"b60c7d3c2f9f73d9d18803467fea6e06a25f372ec62e0b93e29c7896212f99be","abstract_canon_sha256":"a5c104ba0d168bc0fab60ea6a4f5f869b66f8defa914b412a7954b6558dc84ec"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:37:23.764065Z","signature_b64":"FY3ebZ/+vYgM1fF2/y7yXC0jI/tVra3rkVlNIvf0DCPDl8i+mtgmYxIuFWV94bUHA9oMoscaYSYsS6iD3rLKDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dc79c818915d36cb4bac1502aa5e6958bda07b39dd3931c4b92e7ad1655da9ae","last_reissued_at":"2026-07-05T01:37:23.763507Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:37:23.763507Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Manifold Regularization for Locally Stable Deep Neural Networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.LG","cs.NE"],"primary_cat":"stat.ML","authors_text":"Charles Jin, Martin Rinard","submitted_at":"2020-03-09T17:45:44Z","abstract_excerpt":"We apply concepts from manifold regularization to develop new regularization techniques for training locally stable deep neural networks. Our regularizers are based on a sparsification of the graph Laplacian which holds with high probability when the data is sparse in high dimensions, as is common in deep learning. Empirically, our networks exhibit stability in a diverse set of perturbation models, including $\\ell_2$, $\\ell_\\infty$, and Wasserstein-based perturbations; in particular, we achieve 40% adversarial accuracy on CIFAR-10 against an adaptive PGD attack using $\\ell_\\infty$ perturbation"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2003.04286","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/2003.04286/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":"2003.04286","created_at":"2026-07-05T01:37:23.763577+00:00"},{"alias_kind":"arxiv_version","alias_value":"2003.04286v2","created_at":"2026-07-05T01:37:23.763577+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2003.04286","created_at":"2026-07-05T01:37:23.763577+00:00"},{"alias_kind":"pith_short_12","alias_value":"3R44QGERLU3M","created_at":"2026-07-05T01:37:23.763577+00:00"},{"alias_kind":"pith_short_16","alias_value":"3R44QGERLU3MWS5M","created_at":"2026-07-05T01:37:23.763577+00:00"},{"alias_kind":"pith_short_8","alias_value":"3R44QGER","created_at":"2026-07-05T01:37:23.763577+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.04248","citing_title":"Adapting to Evolving Adversaries with Regularized Continual Robust Training","ref_index":33,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3R44QGERLU3MWS5MCUBKUXTJLC","json":"https://pith.science/pith/3R44QGERLU3MWS5MCUBKUXTJLC.json","graph_json":"https://pith.science/api/pith-number/3R44QGERLU3MWS5MCUBKUXTJLC/graph.json","events_json":"https://pith.science/api/pith-number/3R44QGERLU3MWS5MCUBKUXTJLC/events.json","paper":"https://pith.science/paper/3R44QGER"},"agent_actions":{"view_html":"https://pith.science/pith/3R44QGERLU3MWS5MCUBKUXTJLC","download_json":"https://pith.science/pith/3R44QGERLU3MWS5MCUBKUXTJLC.json","view_paper":"https://pith.science/paper/3R44QGER","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2003.04286&json=true","fetch_graph":"https://pith.science/api/pith-number/3R44QGERLU3MWS5MCUBKUXTJLC/graph.json","fetch_events":"https://pith.science/api/pith-number/3R44QGERLU3MWS5MCUBKUXTJLC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3R44QGERLU3MWS5MCUBKUXTJLC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3R44QGERLU3MWS5MCUBKUXTJLC/action/storage_attestation","attest_author":"https://pith.science/pith/3R44QGERLU3MWS5MCUBKUXTJLC/action/author_attestation","sign_citation":"https://pith.science/pith/3R44QGERLU3MWS5MCUBKUXTJLC/action/citation_signature","submit_replication":"https://pith.science/pith/3R44QGERLU3MWS5MCUBKUXTJLC/action/replication_record"}},"created_at":"2026-07-05T01:37:23.763577+00:00","updated_at":"2026-07-05T01:37:23.763577+00:00"}