{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:3MHGN2QP63EP4KTEC2YF2XSQVR","short_pith_number":"pith:3MHGN2QP","schema_version":"1.0","canonical_sha256":"db0e66ea0ff6c8fe2a6416b05d5e50ac53bef15f8a1aeddb7d8c789d06af0520","source":{"kind":"arxiv","id":"2002.07376","version":2},"attestation_state":"computed","paper":{"title":"Picking Winning Tickets Before Training by Preserving Gradient Flow","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Chaoqi Wang, Guodong Zhang, Roger Grosse","submitted_at":"2020-02-18T05:14:47Z","abstract_excerpt":"Overparameterization has been shown to benefit both the optimization and generalization of neural networks, but large networks are resource hungry at both training and test time. Network pruning can reduce test-time resource requirements, but is typically applied to trained networks and therefore cannot avoid the expensive training process. We aim to prune networks at initialization, thereby saving resources at training time as well. Specifically, we argue that efficient training requires preserving the gradient flow through the network. This leads to a simple but effective pruning criterion w"},"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":"2002.07376","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-02-18T05:14:47Z","cross_cats_sorted":["cs.CV","stat.ML"],"title_canon_sha256":"0531d908850e342675514326503dc22825938452b7605b0d750b38bda28ff0ab","abstract_canon_sha256":"200ff7ae661ea673f28eb7ff45d4d030d1ad9ba8838e6e47337d9dca13126498"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:25:24.596470Z","signature_b64":"VRw82Ky+Uee60BO95AW+1hoTM1UGOsmDp/ZQ0CEbQG+1L2rfvJm5Q3UARwkPRMpfN+gZF+RpZ5BnUqoI8JKmDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"db0e66ea0ff6c8fe2a6416b05d5e50ac53bef15f8a1aeddb7d8c789d06af0520","last_reissued_at":"2026-07-05T01:25:24.596038Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:25:24.596038Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Picking Winning Tickets Before Training by Preserving Gradient Flow","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Chaoqi Wang, Guodong Zhang, Roger Grosse","submitted_at":"2020-02-18T05:14:47Z","abstract_excerpt":"Overparameterization has been shown to benefit both the optimization and generalization of neural networks, but large networks are resource hungry at both training and test time. Network pruning can reduce test-time resource requirements, but is typically applied to trained networks and therefore cannot avoid the expensive training process. We aim to prune networks at initialization, thereby saving resources at training time as well. Specifically, we argue that efficient training requires preserving the gradient flow through the network. This leads to a simple but effective pruning criterion w"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2002.07376","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/2002.07376/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":"2002.07376","created_at":"2026-07-05T01:25:24.596099+00:00"},{"alias_kind":"arxiv_version","alias_value":"2002.07376v2","created_at":"2026-07-05T01:25:24.596099+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2002.07376","created_at":"2026-07-05T01:25:24.596099+00:00"},{"alias_kind":"pith_short_12","alias_value":"3MHGN2QP63EP","created_at":"2026-07-05T01:25:24.596099+00:00"},{"alias_kind":"pith_short_16","alias_value":"3MHGN2QP63EP4KTE","created_at":"2026-07-05T01:25:24.596099+00:00"},{"alias_kind":"pith_short_8","alias_value":"3MHGN2QP","created_at":"2026-07-05T01:25:24.596099+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.05478","citing_title":"Can We Predict The Human Preference For Text-to-Image Content Prior To Generation And Is It Even Useful To Do So?","ref_index":66,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30382","citing_title":"RQP: Resource-Oriented Quantiser Pruning for Neural Networks on FPGAs","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2310.12508","citing_title":"SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation","ref_index":170,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12207","citing_title":"Not How Many, But Which: Parameter Placement in Low-Rank Adaptation","ref_index":86,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09988","citing_title":"Engineering Resource-constrained Software Systems with DNN Components: a Concept-based Pruning Approach","ref_index":92,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3MHGN2QP63EP4KTEC2YF2XSQVR","json":"https://pith.science/pith/3MHGN2QP63EP4KTEC2YF2XSQVR.json","graph_json":"https://pith.science/api/pith-number/3MHGN2QP63EP4KTEC2YF2XSQVR/graph.json","events_json":"https://pith.science/api/pith-number/3MHGN2QP63EP4KTEC2YF2XSQVR/events.json","paper":"https://pith.science/paper/3MHGN2QP"},"agent_actions":{"view_html":"https://pith.science/pith/3MHGN2QP63EP4KTEC2YF2XSQVR","download_json":"https://pith.science/pith/3MHGN2QP63EP4KTEC2YF2XSQVR.json","view_paper":"https://pith.science/paper/3MHGN2QP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2002.07376&json=true","fetch_graph":"https://pith.science/api/pith-number/3MHGN2QP63EP4KTEC2YF2XSQVR/graph.json","fetch_events":"https://pith.science/api/pith-number/3MHGN2QP63EP4KTEC2YF2XSQVR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3MHGN2QP63EP4KTEC2YF2XSQVR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3MHGN2QP63EP4KTEC2YF2XSQVR/action/storage_attestation","attest_author":"https://pith.science/pith/3MHGN2QP63EP4KTEC2YF2XSQVR/action/author_attestation","sign_citation":"https://pith.science/pith/3MHGN2QP63EP4KTEC2YF2XSQVR/action/citation_signature","submit_replication":"https://pith.science/pith/3MHGN2QP63EP4KTEC2YF2XSQVR/action/replication_record"}},"created_at":"2026-07-05T01:25:24.596099+00:00","updated_at":"2026-07-05T01:25:24.596099+00:00"}