{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:JY6SRHQLEANQGGJC3ETYJ52YZ3","short_pith_number":"pith:JY6SRHQL","schema_version":"1.0","canonical_sha256":"4e3d289e0b201b031922d92784f758cee331a0890ae0cbef4ee636d0bcc65781","source":{"kind":"arxiv","id":"2202.02931","version":1},"attestation_state":"computed","paper":{"title":"TRGP: Trust Region Gradient Projection for Continual Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Deliang Fan, Junshan Zhang, Li Yang, Sen Lin","submitted_at":"2022-02-07T04:21:54Z","abstract_excerpt":"Catastrophic forgetting is one of the major challenges in continual learning. To address this issue, some existing methods put restrictive constraints on the optimization space of the new task for minimizing the interference to old tasks. However, this may lead to unsatisfactory performance for the new task, especially when the new task is strongly correlated with old tasks. To tackle this challenge, we propose Trust Region Gradient Projection (TRGP) for continual learning to facilitate the forward knowledge transfer based on an efficient characterization of task correlation. Particularly, we "},"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":"2202.02931","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-02-07T04:21:54Z","cross_cats_sorted":[],"title_canon_sha256":"8d8ad560cfeab2230a76b923665099bad0b3f28072753a9981dda8d2d206d402","abstract_canon_sha256":"c5264668e60def8a7351b5974f172652e9223a99e3a83a29e2f06c8520359319"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:54:34.246104Z","signature_b64":"zCED8HUMmQedaRWOKd+4byX0Hvdwkw+1C7IeLanYXtcNrt54P6YNQV9M4hxduIA1lKK2GI/uYIywe+apdpGOAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4e3d289e0b201b031922d92784f758cee331a0890ae0cbef4ee636d0bcc65781","last_reissued_at":"2026-07-05T03:54:34.245673Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:54:34.245673Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TRGP: Trust Region Gradient Projection for Continual Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Deliang Fan, Junshan Zhang, Li Yang, Sen Lin","submitted_at":"2022-02-07T04:21:54Z","abstract_excerpt":"Catastrophic forgetting is one of the major challenges in continual learning. To address this issue, some existing methods put restrictive constraints on the optimization space of the new task for minimizing the interference to old tasks. However, this may lead to unsatisfactory performance for the new task, especially when the new task is strongly correlated with old tasks. To tackle this challenge, we propose Trust Region Gradient Projection (TRGP) for continual learning to facilitate the forward knowledge transfer based on an efficient characterization of task correlation. Particularly, we "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2202.02931","kind":"arxiv","version":1},"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/2202.02931/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":"2202.02931","created_at":"2026-07-05T03:54:34.245733+00:00"},{"alias_kind":"arxiv_version","alias_value":"2202.02931v1","created_at":"2026-07-05T03:54:34.245733+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2202.02931","created_at":"2026-07-05T03:54:34.245733+00:00"},{"alias_kind":"pith_short_12","alias_value":"JY6SRHQLEANQ","created_at":"2026-07-05T03:54:34.245733+00:00"},{"alias_kind":"pith_short_16","alias_value":"JY6SRHQLEANQGGJC","created_at":"2026-07-05T03:54:34.245733+00:00"},{"alias_kind":"pith_short_8","alias_value":"JY6SRHQL","created_at":"2026-07-05T03:54:34.245733+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18528","citing_title":"A Prototypical Signature Approach for Writer-Independent Offline Signature Verification","ref_index":246,"is_internal_anchor":false},{"citing_arxiv_id":"2606.24901","citing_title":"LLM Evolution as an Industry-Scale Ecosystem: A Lifecycle Perspective on Continual Learning","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29164","citing_title":"Invariant Reasoning Directions in Latent Trajectories of Language Models","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20005","citing_title":"Fine-Tuning Without Forgetting via Loss-Adaptive Learning Rates","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11064","citing_title":"A Faster Path to Continual Learning","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JY6SRHQLEANQGGJC3ETYJ52YZ3","json":"https://pith.science/pith/JY6SRHQLEANQGGJC3ETYJ52YZ3.json","graph_json":"https://pith.science/api/pith-number/JY6SRHQLEANQGGJC3ETYJ52YZ3/graph.json","events_json":"https://pith.science/api/pith-number/JY6SRHQLEANQGGJC3ETYJ52YZ3/events.json","paper":"https://pith.science/paper/JY6SRHQL"},"agent_actions":{"view_html":"https://pith.science/pith/JY6SRHQLEANQGGJC3ETYJ52YZ3","download_json":"https://pith.science/pith/JY6SRHQLEANQGGJC3ETYJ52YZ3.json","view_paper":"https://pith.science/paper/JY6SRHQL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2202.02931&json=true","fetch_graph":"https://pith.science/api/pith-number/JY6SRHQLEANQGGJC3ETYJ52YZ3/graph.json","fetch_events":"https://pith.science/api/pith-number/JY6SRHQLEANQGGJC3ETYJ52YZ3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JY6SRHQLEANQGGJC3ETYJ52YZ3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JY6SRHQLEANQGGJC3ETYJ52YZ3/action/storage_attestation","attest_author":"https://pith.science/pith/JY6SRHQLEANQGGJC3ETYJ52YZ3/action/author_attestation","sign_citation":"https://pith.science/pith/JY6SRHQLEANQGGJC3ETYJ52YZ3/action/citation_signature","submit_replication":"https://pith.science/pith/JY6SRHQLEANQGGJC3ETYJ52YZ3/action/replication_record"}},"created_at":"2026-07-05T03:54:34.245733+00:00","updated_at":"2026-07-05T03:54:34.245733+00:00"}