{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:4PYQIVUW2UBUEJUSPS6ARUQ73T","short_pith_number":"pith:4PYQIVUW","schema_version":"1.0","canonical_sha256":"e3f1045696d5034226927cbc08d21fdcd29682d93c0f0b8876fd39af6918d275","source":{"kind":"arxiv","id":"1911.08772","version":1},"attestation_state":"computed","paper":{"title":"Understanding Top-k Sparsification in Distributed Deep Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Ka Chun Cheung, Shaohuai Shi, Simon See, Xiaowen Chu","submitted_at":"2019-11-20T08:50:59Z","abstract_excerpt":"Distributed stochastic gradient descent (SGD) algorithms are widely deployed in training large-scale deep learning models, while the communication overhead among workers becomes the new system bottleneck. Recently proposed gradient sparsification techniques, especially Top-$k$ sparsification with error compensation (TopK-SGD), can significantly reduce the communication traffic without an obvious impact on the model accuracy. Some theoretical studies have been carried out to analyze the convergence property of TopK-SGD. However, existing studies do not dive into the details of Top-$k$ operator "},"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":"1911.08772","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-11-20T08:50:59Z","cross_cats_sorted":["cs.DC","stat.ML"],"title_canon_sha256":"25d14fc1ebcdc0bfa12c83db11fecbcfb48bca8a936c696e013a0ebe94cb48ed","abstract_canon_sha256":"32ec828738853c78a75644a6ca268994f8edffa2dbe5c03257988c2e26a8b44b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T00:20:45.366064Z","signature_b64":"RJgM/vm4JwI3fgIhAPunzNnlboB0s245J/9FIKh11H3whGenJRUGWgtAfe+d/uafWZ3vj8h6V1y9i4/vn0/+CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e3f1045696d5034226927cbc08d21fdcd29682d93c0f0b8876fd39af6918d275","last_reissued_at":"2026-07-05T00:20:45.365558Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T00:20:45.365558Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Understanding Top-k Sparsification in Distributed Deep Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Ka Chun Cheung, Shaohuai Shi, Simon See, Xiaowen Chu","submitted_at":"2019-11-20T08:50:59Z","abstract_excerpt":"Distributed stochastic gradient descent (SGD) algorithms are widely deployed in training large-scale deep learning models, while the communication overhead among workers becomes the new system bottleneck. Recently proposed gradient sparsification techniques, especially Top-$k$ sparsification with error compensation (TopK-SGD), can significantly reduce the communication traffic without an obvious impact on the model accuracy. Some theoretical studies have been carried out to analyze the convergence property of TopK-SGD. However, existing studies do not dive into the details of Top-$k$ operator "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1911.08772","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/1911.08772/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":"1911.08772","created_at":"2026-07-05T00:20:45.365619+00:00"},{"alias_kind":"arxiv_version","alias_value":"1911.08772v1","created_at":"2026-07-05T00:20:45.365619+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1911.08772","created_at":"2026-07-05T00:20:45.365619+00:00"},{"alias_kind":"pith_short_12","alias_value":"4PYQIVUW2UBU","created_at":"2026-07-05T00:20:45.365619+00:00"},{"alias_kind":"pith_short_16","alias_value":"4PYQIVUW2UBUEJUS","created_at":"2026-07-05T00:20:45.365619+00:00"},{"alias_kind":"pith_short_8","alias_value":"4PYQIVUW","created_at":"2026-07-05T00:20:45.365619+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.27844","citing_title":"ZipCCL: Efficient Lossless Data Compression of Communication Collectives for Accelerating LLM Training","ref_index":41,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4PYQIVUW2UBUEJUSPS6ARUQ73T","json":"https://pith.science/pith/4PYQIVUW2UBUEJUSPS6ARUQ73T.json","graph_json":"https://pith.science/api/pith-number/4PYQIVUW2UBUEJUSPS6ARUQ73T/graph.json","events_json":"https://pith.science/api/pith-number/4PYQIVUW2UBUEJUSPS6ARUQ73T/events.json","paper":"https://pith.science/paper/4PYQIVUW"},"agent_actions":{"view_html":"https://pith.science/pith/4PYQIVUW2UBUEJUSPS6ARUQ73T","download_json":"https://pith.science/pith/4PYQIVUW2UBUEJUSPS6ARUQ73T.json","view_paper":"https://pith.science/paper/4PYQIVUW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1911.08772&json=true","fetch_graph":"https://pith.science/api/pith-number/4PYQIVUW2UBUEJUSPS6ARUQ73T/graph.json","fetch_events":"https://pith.science/api/pith-number/4PYQIVUW2UBUEJUSPS6ARUQ73T/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4PYQIVUW2UBUEJUSPS6ARUQ73T/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4PYQIVUW2UBUEJUSPS6ARUQ73T/action/storage_attestation","attest_author":"https://pith.science/pith/4PYQIVUW2UBUEJUSPS6ARUQ73T/action/author_attestation","sign_citation":"https://pith.science/pith/4PYQIVUW2UBUEJUSPS6ARUQ73T/action/citation_signature","submit_replication":"https://pith.science/pith/4PYQIVUW2UBUEJUSPS6ARUQ73T/action/replication_record"}},"created_at":"2026-07-05T00:20:45.365619+00:00","updated_at":"2026-07-05T00:20:45.365619+00:00"}