{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2017:N672T5PKFSHVYHVOUHCFSVLRM3","short_pith_number":"pith:N672T5PK","schema_version":"1.0","canonical_sha256":"6fbfa9f5ea2c8f5c1eaea1c459557166fed2cc0694755fef4aaab9ba89c051a7","source":{"kind":"arxiv","id":"1705.07878","version":6},"attestation_state":"computed","paper":{"title":"TernGrad: Ternary Gradients to Reduce Communication in Distributed Deep Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","cs.NE"],"primary_cat":"cs.LG","authors_text":"Chunpeng Wu, Cong Xu, Feng Yan, Hai Li, Wei Wen, Yandan Wang, Yiran Chen","submitted_at":"2017-05-22T17:42:15Z","abstract_excerpt":"High network communication cost for synchronizing gradients and parameters is the well-known bottleneck of distributed training. In this work, we propose TernGrad that uses ternary gradients to accelerate distributed deep learning in data parallelism. Our approach requires only three numerical levels {-1,0,1}, which can aggressively reduce the communication time. We mathematically prove the convergence of TernGrad under the assumption of a bound on gradients. Guided by the bound, we propose layer-wise ternarizing and gradient clipping to improve its convergence. Our experiments show that apply"},"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":"1705.07878","kind":"arxiv","version":6},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2017-05-22T17:42:15Z","cross_cats_sorted":["cs.DC","cs.NE"],"title_canon_sha256":"0ca75976f56ec846cc8fc3f04070945df2f541780e91cb83abc34bf5bef52e43","abstract_canon_sha256":"ad15118648d3ba3858517b72b4309eb7c61661e66fd8f0b8ea21a9066123d7f7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:27:04.651938Z","signature_b64":"kb/IXFN1hDUA/+m8kiy3Cw4mK5u/lDBkoDZj3jHNBRx24b1ciNR6w9xp/m1XiSwq6lxoSh/NIKM+ioAKU+40Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6fbfa9f5ea2c8f5c1eaea1c459557166fed2cc0694755fef4aaab9ba89c051a7","last_reissued_at":"2026-05-18T00:27:04.651423Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:27:04.651423Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TernGrad: Ternary Gradients to Reduce Communication in Distributed Deep Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","cs.NE"],"primary_cat":"cs.LG","authors_text":"Chunpeng Wu, Cong Xu, Feng Yan, Hai Li, Wei Wen, Yandan Wang, Yiran Chen","submitted_at":"2017-05-22T17:42:15Z","abstract_excerpt":"High network communication cost for synchronizing gradients and parameters is the well-known bottleneck of distributed training. In this work, we propose TernGrad that uses ternary gradients to accelerate distributed deep learning in data parallelism. Our approach requires only three numerical levels {-1,0,1}, which can aggressively reduce the communication time. We mathematically prove the convergence of TernGrad under the assumption of a bound on gradients. Guided by the bound, we propose layer-wise ternarizing and gradient clipping to improve its convergence. Our experiments show that apply"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1705.07878","kind":"arxiv","version":6},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"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":"1705.07878","created_at":"2026-05-18T00:27:04.651497+00:00"},{"alias_kind":"arxiv_version","alias_value":"1705.07878v6","created_at":"2026-05-18T00:27:04.651497+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1705.07878","created_at":"2026-05-18T00:27:04.651497+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.25550","citing_title":"Enhancing SignSGD: Small-Batch Convergence Analysis and a Hybrid Switching Strategy","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/N672T5PKFSHVYHVOUHCFSVLRM3","json":"https://pith.science/pith/N672T5PKFSHVYHVOUHCFSVLRM3.json","graph_json":"https://pith.science/api/pith-number/N672T5PKFSHVYHVOUHCFSVLRM3/graph.json","events_json":"https://pith.science/api/pith-number/N672T5PKFSHVYHVOUHCFSVLRM3/events.json","paper":"https://pith.science/paper/N672T5PK"},"agent_actions":{"view_html":"https://pith.science/pith/N672T5PKFSHVYHVOUHCFSVLRM3","download_json":"https://pith.science/pith/N672T5PKFSHVYHVOUHCFSVLRM3.json","view_paper":"https://pith.science/paper/N672T5PK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1705.07878&json=true","fetch_graph":"https://pith.science/api/pith-number/N672T5PKFSHVYHVOUHCFSVLRM3/graph.json","fetch_events":"https://pith.science/api/pith-number/N672T5PKFSHVYHVOUHCFSVLRM3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/N672T5PKFSHVYHVOUHCFSVLRM3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/N672T5PKFSHVYHVOUHCFSVLRM3/action/storage_attestation","attest_author":"https://pith.science/pith/N672T5PKFSHVYHVOUHCFSVLRM3/action/author_attestation","sign_citation":"https://pith.science/pith/N672T5PKFSHVYHVOUHCFSVLRM3/action/citation_signature","submit_replication":"https://pith.science/pith/N672T5PKFSHVYHVOUHCFSVLRM3/action/replication_record"}},"created_at":"2026-05-18T00:27:04.651497+00:00","updated_at":"2026-05-18T00:27:04.651497+00:00"}