{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:PNFPHYOLD4J4NKHPGPJR4NIXHC","short_pith_number":"pith:PNFPHYOL","schema_version":"1.0","canonical_sha256":"7b4af3e1cb1f13c6a8ef33d31e351738afcc6fec5701147bce4d895eecd08bfd","source":{"kind":"arxiv","id":"2108.11179","version":2},"attestation_state":"computed","paper":{"title":"Recall@k Surrogate Loss with Large Batches and Similarity Mixup","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Giorgos Tolias, Jiri Matas, Yash Patel","submitted_at":"2021-08-25T11:09:11Z","abstract_excerpt":"This work focuses on learning deep visual representation models for retrieval by exploring the interplay between a new loss function, the batch size, and a new regularization approach. Direct optimization, by gradient descent, of an evaluation metric, is not possible when it is non-differentiable, which is the case for recall in retrieval. A differentiable surrogate loss for the recall is proposed in this work. Using an implementation that sidesteps the hardware constraints of the GPU memory, the method trains with a very large batch size, which is essential for metrics computed on the entire "},"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":"2108.11179","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2021-08-25T11:09:11Z","cross_cats_sorted":[],"title_canon_sha256":"60cc6fdba74015dd174c7ef291cf584e698299ebd2c643dc446554a8a2468d3f","abstract_canon_sha256":"6618a23f795b70c0c9dab6e825d580ace88a4ccc279b150c9406d8687b2051bb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:08:25.185521Z","signature_b64":"3gJ9ofJt0nULY9Lm6d6CWx7vOjsJ0evIm4hn688XbavOMjglPoP6K+phN+DU3btxGn9sO4faYDwUajo1fTyKAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7b4af3e1cb1f13c6a8ef33d31e351738afcc6fec5701147bce4d895eecd08bfd","last_reissued_at":"2026-07-05T04:08:25.184949Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:08:25.184949Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Recall@k Surrogate Loss with Large Batches and Similarity Mixup","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Giorgos Tolias, Jiri Matas, Yash Patel","submitted_at":"2021-08-25T11:09:11Z","abstract_excerpt":"This work focuses on learning deep visual representation models for retrieval by exploring the interplay between a new loss function, the batch size, and a new regularization approach. Direct optimization, by gradient descent, of an evaluation metric, is not possible when it is non-differentiable, which is the case for recall in retrieval. A differentiable surrogate loss for the recall is proposed in this work. Using an implementation that sidesteps the hardware constraints of the GPU memory, the method trains with a very large batch size, which is essential for metrics computed on the entire "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2108.11179","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/2108.11179/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":"2108.11179","created_at":"2026-07-05T04:08:25.185009+00:00"},{"alias_kind":"arxiv_version","alias_value":"2108.11179v2","created_at":"2026-07-05T04:08:25.185009+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2108.11179","created_at":"2026-07-05T04:08:25.185009+00:00"},{"alias_kind":"pith_short_12","alias_value":"PNFPHYOLD4J4","created_at":"2026-07-05T04:08:25.185009+00:00"},{"alias_kind":"pith_short_16","alias_value":"PNFPHYOLD4J4NKHP","created_at":"2026-07-05T04:08:25.185009+00:00"},{"alias_kind":"pith_short_8","alias_value":"PNFPHYOL","created_at":"2026-07-05T04:08:25.185009+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19684","citing_title":"Exploring Multi-Modal Large Language Models and Two-Stage Fine-Tuning for Fashion Image Retrieval","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PNFPHYOLD4J4NKHPGPJR4NIXHC","json":"https://pith.science/pith/PNFPHYOLD4J4NKHPGPJR4NIXHC.json","graph_json":"https://pith.science/api/pith-number/PNFPHYOLD4J4NKHPGPJR4NIXHC/graph.json","events_json":"https://pith.science/api/pith-number/PNFPHYOLD4J4NKHPGPJR4NIXHC/events.json","paper":"https://pith.science/paper/PNFPHYOL"},"agent_actions":{"view_html":"https://pith.science/pith/PNFPHYOLD4J4NKHPGPJR4NIXHC","download_json":"https://pith.science/pith/PNFPHYOLD4J4NKHPGPJR4NIXHC.json","view_paper":"https://pith.science/paper/PNFPHYOL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2108.11179&json=true","fetch_graph":"https://pith.science/api/pith-number/PNFPHYOLD4J4NKHPGPJR4NIXHC/graph.json","fetch_events":"https://pith.science/api/pith-number/PNFPHYOLD4J4NKHPGPJR4NIXHC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PNFPHYOLD4J4NKHPGPJR4NIXHC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PNFPHYOLD4J4NKHPGPJR4NIXHC/action/storage_attestation","attest_author":"https://pith.science/pith/PNFPHYOLD4J4NKHPGPJR4NIXHC/action/author_attestation","sign_citation":"https://pith.science/pith/PNFPHYOLD4J4NKHPGPJR4NIXHC/action/citation_signature","submit_replication":"https://pith.science/pith/PNFPHYOLD4J4NKHPGPJR4NIXHC/action/replication_record"}},"created_at":"2026-07-05T04:08:25.185009+00:00","updated_at":"2026-07-05T04:08:25.185009+00:00"}