{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:EOPJYU7B66WHWMJH2WEQOD2IBD","short_pith_number":"pith:EOPJYU7B","schema_version":"1.0","canonical_sha256":"239e9c53e1f7ac7b3127d589070f4808ccc15c8f1826573fcc74fe8690c7c7c4","source":{"kind":"arxiv","id":"2406.01755","version":1},"attestation_state":"computed","paper":{"title":"Sparser, Better, Deeper, Stronger: Improving Sparse Training with Exact Orthogonal Initialization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aleksandra Irena Nowak, Filip Szatkowski, Jacek Tabor, {\\L}ukasz Gniecki","submitted_at":"2024-06-03T19:44:47Z","abstract_excerpt":"Static sparse training aims to train sparse models from scratch, achieving remarkable results in recent years. A key design choice is given by the sparse initialization, which determines the trainable sub-network through a binary mask. Existing methods mainly select such mask based on a predefined dense initialization. Such an approach may not efficiently leverage the mask's potential impact on the optimization. An alternative direction, inspired by research into dynamical isometry, is to introduce orthogonality in the sparse subnetwork, which helps in stabilizing the gradient signal. In this "},"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":"2406.01755","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-03T19:44:47Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"1bd16c436c7d7e02bc1f21d837f69a4b58913af81c821427bdbe0e5c87b85ddd","abstract_canon_sha256":"7c6c46292233f48f90029885ac3d60559e2d2dc3c85dc5b7be74b9f15c9acd9f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:26:48.062182Z","signature_b64":"A7nQX7qouCumqiBpkLwWxPhtcIgX6UEwjbylGk4C7XhEn8HYzrSDoSFs9XCtwSEZ8aqza4xAlHjlEmaU2+j8Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"239e9c53e1f7ac7b3127d589070f4808ccc15c8f1826573fcc74fe8690c7c7c4","last_reissued_at":"2026-07-05T08:26:48.061706Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:26:48.061706Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sparser, Better, Deeper, Stronger: Improving Sparse Training with Exact Orthogonal Initialization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Aleksandra Irena Nowak, Filip Szatkowski, Jacek Tabor, {\\L}ukasz Gniecki","submitted_at":"2024-06-03T19:44:47Z","abstract_excerpt":"Static sparse training aims to train sparse models from scratch, achieving remarkable results in recent years. A key design choice is given by the sparse initialization, which determines the trainable sub-network through a binary mask. Existing methods mainly select such mask based on a predefined dense initialization. Such an approach may not efficiently leverage the mask's potential impact on the optimization. An alternative direction, inspired by research into dynamical isometry, is to introduce orthogonality in the sparse subnetwork, which helps in stabilizing the gradient signal. In this "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.01755","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/2406.01755/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":"2406.01755","created_at":"2026-07-05T08:26:48.061763+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.01755v1","created_at":"2026-07-05T08:26:48.061763+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.01755","created_at":"2026-07-05T08:26:48.061763+00:00"},{"alias_kind":"pith_short_12","alias_value":"EOPJYU7B66WH","created_at":"2026-07-05T08:26:48.061763+00:00"},{"alias_kind":"pith_short_16","alias_value":"EOPJYU7B66WHWMJH","created_at":"2026-07-05T08:26:48.061763+00:00"},{"alias_kind":"pith_short_8","alias_value":"EOPJYU7B","created_at":"2026-07-05T08:26:48.061763+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.10560","citing_title":"Heterogeneous Connectivity in Sparse Networks: Fan-in Profiles, Gradient Hierarchy, and Topological Equilibria","ref_index":32,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EOPJYU7B66WHWMJH2WEQOD2IBD","json":"https://pith.science/pith/EOPJYU7B66WHWMJH2WEQOD2IBD.json","graph_json":"https://pith.science/api/pith-number/EOPJYU7B66WHWMJH2WEQOD2IBD/graph.json","events_json":"https://pith.science/api/pith-number/EOPJYU7B66WHWMJH2WEQOD2IBD/events.json","paper":"https://pith.science/paper/EOPJYU7B"},"agent_actions":{"view_html":"https://pith.science/pith/EOPJYU7B66WHWMJH2WEQOD2IBD","download_json":"https://pith.science/pith/EOPJYU7B66WHWMJH2WEQOD2IBD.json","view_paper":"https://pith.science/paper/EOPJYU7B","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.01755&json=true","fetch_graph":"https://pith.science/api/pith-number/EOPJYU7B66WHWMJH2WEQOD2IBD/graph.json","fetch_events":"https://pith.science/api/pith-number/EOPJYU7B66WHWMJH2WEQOD2IBD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EOPJYU7B66WHWMJH2WEQOD2IBD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EOPJYU7B66WHWMJH2WEQOD2IBD/action/storage_attestation","attest_author":"https://pith.science/pith/EOPJYU7B66WHWMJH2WEQOD2IBD/action/author_attestation","sign_citation":"https://pith.science/pith/EOPJYU7B66WHWMJH2WEQOD2IBD/action/citation_signature","submit_replication":"https://pith.science/pith/EOPJYU7B66WHWMJH2WEQOD2IBD/action/replication_record"}},"created_at":"2026-07-05T08:26:48.061763+00:00","updated_at":"2026-07-05T08:26:48.061763+00:00"}