{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:F3BV7YOY2BTMR5QEIWMVB6T5GG","short_pith_number":"pith:F3BV7YOY","schema_version":"1.0","canonical_sha256":"2ec35fe1d8d066c8f604459950fa7d318d3059c2a6dac117d33846c0c832a2cc","source":{"kind":"arxiv","id":"1802.08021","version":3},"attestation_state":"computed","paper":{"title":"SparCML: High-Performance Sparse Communication for Machine Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.DC","authors_text":"Cedric Renggli, Dan Alistarh, Mehdi Aghagolzadeh, Saleh Ashkboos, Torsten Hoefler","submitted_at":"2018-02-22T13:02:53Z","abstract_excerpt":"Applying machine learning techniques to the quickly growing data in science and industry requires highly-scalable algorithms. Large datasets are most commonly processed \"data parallel\" distributed across many nodes. Each node's contribution to the overall gradient is summed using a global allreduce. This allreduce is the single communication and thus scalability bottleneck for most machine learning workloads. We observe that frequently, many gradient values are (close to) zero, leading to sparse of sparsifyable communications. To exploit this insight, we analyze, design, and implement a set of"},"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":"1802.08021","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2018-02-22T13:02:53Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"508b3b1459fb4ed863efd102a88eb157925da04c4e2fdefa29d12614755c9c6f","abstract_canon_sha256":"58c2ce834e89983d741c58dd2e8d9be91f2ef65369809115123e4a9c01c8770c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-04T23:57:48.777581Z","signature_b64":"3hyT+zCRGf0U5iUGmjF/QWSR9u1xwzE8FkF0u1F/7RKV7INLT9O0sFLFMO6+svRXtLAzenYWvci+XR1RXxBoCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2ec35fe1d8d066c8f604459950fa7d318d3059c2a6dac117d33846c0c832a2cc","last_reissued_at":"2026-07-04T23:57:48.777171Z","signature_status":"signed_v1","first_computed_at":"2026-07-04T23:57:48.777171Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SparCML: High-Performance Sparse Communication for Machine Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.DC","authors_text":"Cedric Renggli, Dan Alistarh, Mehdi Aghagolzadeh, Saleh Ashkboos, Torsten Hoefler","submitted_at":"2018-02-22T13:02:53Z","abstract_excerpt":"Applying machine learning techniques to the quickly growing data in science and industry requires highly-scalable algorithms. Large datasets are most commonly processed \"data parallel\" distributed across many nodes. Each node's contribution to the overall gradient is summed using a global allreduce. This allreduce is the single communication and thus scalability bottleneck for most machine learning workloads. We observe that frequently, many gradient values are (close to) zero, leading to sparse of sparsifyable communications. To exploit this insight, we analyze, design, and implement a set of"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1802.08021","kind":"arxiv","version":3},"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/1802.08021/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":"1802.08021","created_at":"2026-07-04T23:57:48.777226+00:00"},{"alias_kind":"arxiv_version","alias_value":"1802.08021v3","created_at":"2026-07-04T23:57:48.777226+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1802.08021","created_at":"2026-07-04T23:57:48.777226+00:00"},{"alias_kind":"pith_short_12","alias_value":"F3BV7YOY2BTM","created_at":"2026-07-04T23:57:48.777226+00:00"},{"alias_kind":"pith_short_16","alias_value":"F3BV7YOY2BTMR5QE","created_at":"2026-07-04T23:57:48.777226+00:00"},{"alias_kind":"pith_short_8","alias_value":"F3BV7YOY","created_at":"2026-07-04T23:57:48.777226+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"1908.04207","citing_title":"Taming Unbalanced Training Workloads in Deep Learning with Partial Collective Operations","ref_index":47,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/F3BV7YOY2BTMR5QEIWMVB6T5GG","json":"https://pith.science/pith/F3BV7YOY2BTMR5QEIWMVB6T5GG.json","graph_json":"https://pith.science/api/pith-number/F3BV7YOY2BTMR5QEIWMVB6T5GG/graph.json","events_json":"https://pith.science/api/pith-number/F3BV7YOY2BTMR5QEIWMVB6T5GG/events.json","paper":"https://pith.science/paper/F3BV7YOY"},"agent_actions":{"view_html":"https://pith.science/pith/F3BV7YOY2BTMR5QEIWMVB6T5GG","download_json":"https://pith.science/pith/F3BV7YOY2BTMR5QEIWMVB6T5GG.json","view_paper":"https://pith.science/paper/F3BV7YOY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1802.08021&json=true","fetch_graph":"https://pith.science/api/pith-number/F3BV7YOY2BTMR5QEIWMVB6T5GG/graph.json","fetch_events":"https://pith.science/api/pith-number/F3BV7YOY2BTMR5QEIWMVB6T5GG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F3BV7YOY2BTMR5QEIWMVB6T5GG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F3BV7YOY2BTMR5QEIWMVB6T5GG/action/storage_attestation","attest_author":"https://pith.science/pith/F3BV7YOY2BTMR5QEIWMVB6T5GG/action/author_attestation","sign_citation":"https://pith.science/pith/F3BV7YOY2BTMR5QEIWMVB6T5GG/action/citation_signature","submit_replication":"https://pith.science/pith/F3BV7YOY2BTMR5QEIWMVB6T5GG/action/replication_record"}},"created_at":"2026-07-04T23:57:48.777226+00:00","updated_at":"2026-07-04T23:57:48.777226+00:00"}