{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2016:55UOMO4HJXQO3AEGMJRMZ3JF4Y","short_pith_number":"pith:55UOMO4H","canonical_record":{"source":{"id":"1610.05492","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2016-10-18T09:11:51Z","cross_cats_sorted":[],"title_canon_sha256":"e5b84b937bf0a2a632f1918628df4d3bcd260628291c3c43ae959b77d377f37d","abstract_canon_sha256":"3e7f111c4100d8648a1201de071b7bd64fce5e7e5e8239aff271f4c81380deab"},"schema_version":"1.0"},"canonical_sha256":"ef68e63b874de0ed80866262cced25e6096ba6b72bbc20af874be34d84f1077f","source":{"kind":"arxiv","id":"1610.05492","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"1610.05492","created_at":"2026-07-04T22:18:31Z"},{"alias_kind":"arxiv_version","alias_value":"1610.05492v2","created_at":"2026-07-04T22:18:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1610.05492","created_at":"2026-07-04T22:18:31Z"},{"alias_kind":"pith_short_12","alias_value":"55UOMO4HJXQO","created_at":"2026-07-04T22:18:31Z"},{"alias_kind":"pith_short_16","alias_value":"55UOMO4HJXQO3AEG","created_at":"2026-07-04T22:18:31Z"},{"alias_kind":"pith_short_8","alias_value":"55UOMO4H","created_at":"2026-07-04T22:18:31Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2016:55UOMO4HJXQO3AEGMJRMZ3JF4Y","target":"record","payload":{"canonical_record":{"source":{"id":"1610.05492","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2016-10-18T09:11:51Z","cross_cats_sorted":[],"title_canon_sha256":"e5b84b937bf0a2a632f1918628df4d3bcd260628291c3c43ae959b77d377f37d","abstract_canon_sha256":"3e7f111c4100d8648a1201de071b7bd64fce5e7e5e8239aff271f4c81380deab"},"schema_version":"1.0"},"canonical_sha256":"ef68e63b874de0ed80866262cced25e6096ba6b72bbc20af874be34d84f1077f","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-04T22:18:31.926054Z","signature_b64":"2gYtBRspNBHdOnDVOi776SmevSX/ciNeyLK+iMThUeLqQ45DPGpQzCl7skoM54qPXATo/FLNArcbefVdOX1FBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ef68e63b874de0ed80866262cced25e6096ba6b72bbc20af874be34d84f1077f","last_reissued_at":"2026-07-04T22:18:31.925547Z","signature_status":"signed_v1","first_computed_at":"2026-07-04T22:18:31.925547Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"1610.05492","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-04T22:18:31Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"FjgMGiSuduwcIjTVF1gBoojcnOi8SiIsHQ8l5DiyAUQ3mIAs7WDLg/yMZIpIXxr3+eUD53gDj6nUCpG/zHJ6BQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-19T09:42:47.514050Z"},"content_sha256":"127a4df53ba0021b7d84c237b4cf05ff6218e4e60062ad6956bbee8764f37494","schema_version":"1.0","event_id":"sha256:127a4df53ba0021b7d84c237b4cf05ff6218e4e60062ad6956bbee8764f37494"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2016:55UOMO4HJXQO3AEGMJRMZ3JF4Y","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Federated Learning: Strategies for Improving Communication Efficiency","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Federated learning trains high-quality models on mobile devices while reducing uplink communication by up to 100 times.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ananda Theertha Suresh, Dave Bacon, Felix X. Yu, H. Brendan McMahan, Jakub Kone\\v{c}n\\'y, Peter Richt\\'arik","submitted_at":"2016-10-18T09:11:51Z","abstract_excerpt":"Federated Learning is a machine learning setting where the goal is to train a high-quality centralized model while training data remains distributed over a large number of clients each with unreliable and relatively slow network connections. We consider learning algorithms for this setting where on each round, each client independently computes an update to the current model based on its local data, and communicates this update to a central server, where the client-side updates are aggregated to compute a new global model. The typical clients in this setting are mobile phones, and communicatio"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on both convolutional and recurrent networks show that the proposed methods can reduce the communication cost by two orders of magnitude.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That restricting updates to a low-rank or masked space or compressing them via quantization and subsampling preserves sufficient information for the aggregated global model to converge to high quality.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Structured updates (low-rank or masked) and sketched updates (quantized, rotated, subsampled) reduce uplink communication in federated learning by up to two orders of magnitude on convolutional and recurrent networks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Federated learning trains high-quality models on mobile devices while reducing uplink communication by up to 100 times.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"d6c2708cb8d2039dd365294c109ab5908aba116570f7618516c51bba450e5814"},"source":{"id":"1610.05492","kind":"arxiv","version":2},"verdict":{"id":"5848767c-a797-431e-8e33-1dee5d14ec48","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T13:38:18.507126Z","strongest_claim":"Experiments on both convolutional and recurrent networks show that the proposed methods can reduce the communication cost by two orders of magnitude.","one_line_summary":"Structured updates (low-rank or masked) and sketched updates (quantized, rotated, subsampled) reduce uplink communication in federated learning by up to two orders of magnitude on convolutional and recurrent networks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That restricting updates to a low-rank or masked space or compressing them via quantization and subsampling preserves sufficient information for the aggregated global model to converge to high quality.","pith_extraction_headline":"Federated learning trains high-quality models on mobile devices while reducing uplink communication by up to 100 times."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/1610.05492/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":25,"sample":[{"doi":"","year":2016,"title":"Conversational contextual cues: The case of personalization and history for response ranking","work_id":"d3e65bc8-ddf0-42d9-8ebb-cc1d967d5721","ref_index":1,"cited_arxiv_id":"1606.00372","is_internal_anchor":true},{"doi":"","year":2016,"title":"QSGD: Communication-Efficient SGD via Gradient Quantization and Encoding","work_id":"cac53f8a-ecd0-4a1e-b6e0-0249417a3593","ref_index":2,"cited_arxiv_id":"1610.02132","is_internal_anchor":false},{"doi":"","year":2017,"title":"Brendan McMahan, Sarvar Patel, Daniel Ramage, Aaron Segal, and Karn Seth","work_id":"e6ed76b0-7305-4f98-8e60-2f85232cc5a4","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2014,"title":"Project adam: Building an efficient and scalable deep learning training system","work_id":"4aaeb4dc-8127-41b8-a527-561004a6aaa4","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2012,"title":"Large scale distributed deep networks","work_id":"6ecb0dd1-c484-4b8f-bd82-1fddb78685b2","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":25,"snapshot_sha256":"6a965039656dcbabe8ba0c7c87c3734ae4b2b0c63e667638758e223711e80b0d","internal_anchors":5},"formal_canon":{"evidence_count":1,"snapshot_sha256":"a95dfb44f62d940491a5f6ee47b152d204fa1402e1775895e8bcd6b23116cd37"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"5848767c-a797-431e-8e33-1dee5d14ec48"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-04T22:18:31Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"0ruKjPicop25OzJoodDHWQI1UUtXkoNG1WVYoLsV6EXxcuICkqPjR0q9XRzDP9oeglWTvvXsRTC9dztnL3kUAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-19T09:42:47.515405Z"},"content_sha256":"b63e08f7576fe093d1ef0198c684ca8f1e0af09d8beffe37d4407159d78c0829","schema_version":"1.0","event_id":"sha256:b63e08f7576fe093d1ef0198c684ca8f1e0af09d8beffe37d4407159d78c0829"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/55UOMO4HJXQO3AEGMJRMZ3JF4Y/bundle.json","state_url":"https://pith.science/pith/55UOMO4HJXQO3AEGMJRMZ3JF4Y/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/55UOMO4HJXQO3AEGMJRMZ3JF4Y/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-19T09:42:47Z","links":{"resolver":"https://pith.science/pith/55UOMO4HJXQO3AEGMJRMZ3JF4Y","bundle":"https://pith.science/pith/55UOMO4HJXQO3AEGMJRMZ3JF4Y/bundle.json","state":"https://pith.science/pith/55UOMO4HJXQO3AEGMJRMZ3JF4Y/state.json","well_known_bundle":"https://pith.science/.well-known/pith/55UOMO4HJXQO3AEGMJRMZ3JF4Y/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2016:55UOMO4HJXQO3AEGMJRMZ3JF4Y","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"3e7f111c4100d8648a1201de071b7bd64fce5e7e5e8239aff271f4c81380deab","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2016-10-18T09:11:51Z","title_canon_sha256":"e5b84b937bf0a2a632f1918628df4d3bcd260628291c3c43ae959b77d377f37d"},"schema_version":"1.0","source":{"id":"1610.05492","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"1610.05492","created_at":"2026-07-04T22:18:31Z"},{"alias_kind":"arxiv_version","alias_value":"1610.05492v2","created_at":"2026-07-04T22:18:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1610.05492","created_at":"2026-07-04T22:18:31Z"},{"alias_kind":"pith_short_12","alias_value":"55UOMO4HJXQO","created_at":"2026-07-04T22:18:31Z"},{"alias_kind":"pith_short_16","alias_value":"55UOMO4HJXQO3AEG","created_at":"2026-07-04T22:18:31Z"},{"alias_kind":"pith_short_8","alias_value":"55UOMO4H","created_at":"2026-07-04T22:18:31Z"}],"graph_snapshots":[{"event_id":"sha256:b63e08f7576fe093d1ef0198c684ca8f1e0af09d8beffe37d4407159d78c0829","target":"graph","created_at":"2026-07-04T22:18:31Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Experiments on both convolutional and recurrent networks show that the proposed methods can reduce the communication cost by two orders of magnitude."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That restricting updates to a low-rank or masked space or compressing them via quantization and subsampling preserves sufficient information for the aggregated global model to converge to high quality."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Structured updates (low-rank or masked) and sketched updates (quantized, rotated, subsampled) reduce uplink communication in federated learning by up to two orders of magnitude on convolutional and recurrent networks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Federated learning trains high-quality models on mobile devices while reducing uplink communication by up to 100 times."}],"snapshot_sha256":"d6c2708cb8d2039dd365294c109ab5908aba116570f7618516c51bba450e5814"},"formal_canon":{"evidence_count":1,"snapshot_sha256":"a95dfb44f62d940491a5f6ee47b152d204fa1402e1775895e8bcd6b23116cd37"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/1610.05492/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Federated Learning is a machine learning setting where the goal is to train a high-quality centralized model while training data remains distributed over a large number of clients each with unreliable and relatively slow network connections. We consider learning algorithms for this setting where on each round, each client independently computes an update to the current model based on its local data, and communicates this update to a central server, where the client-side updates are aggregated to compute a new global model. The typical clients in this setting are mobile phones, and communicatio","authors_text":"Ananda Theertha Suresh, Dave Bacon, Felix X. Yu, H. Brendan McMahan, Jakub Kone\\v{c}n\\'y, Peter Richt\\'arik","cross_cats":[],"headline":"Federated learning trains high-quality models on mobile devices while reducing uplink communication by up to 100 times.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2016-10-18T09:11:51Z","title":"Federated Learning: Strategies for Improving Communication Efficiency"},"references":{"count":25,"internal_anchors":5,"resolved_work":25,"sample":[{"cited_arxiv_id":"1606.00372","doi":"","is_internal_anchor":true,"ref_index":1,"title":"Conversational contextual cues: The case of personalization and history for response ranking","work_id":"d3e65bc8-ddf0-42d9-8ebb-cc1d967d5721","year":2016},{"cited_arxiv_id":"1610.02132","doi":"","is_internal_anchor":false,"ref_index":2,"title":"QSGD: Communication-Efficient SGD via Gradient Quantization and Encoding","work_id":"cac53f8a-ecd0-4a1e-b6e0-0249417a3593","year":2016},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Brendan McMahan, Sarvar Patel, Daniel Ramage, Aaron Segal, and Karn Seth","work_id":"e6ed76b0-7305-4f98-8e60-2f85232cc5a4","year":2017},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Project adam: Building an efficient and scalable deep learning training system","work_id":"4aaeb4dc-8127-41b8-a527-561004a6aaa4","year":2014},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Large scale distributed deep networks","work_id":"6ecb0dd1-c484-4b8f-bd82-1fddb78685b2","year":2012}],"snapshot_sha256":"6a965039656dcbabe8ba0c7c87c3734ae4b2b0c63e667638758e223711e80b0d"},"source":{"id":"1610.05492","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-12T13:38:18.507126Z","id":"5848767c-a797-431e-8e33-1dee5d14ec48","model_set":{"reader":"grok-4.3"},"one_line_summary":"Structured updates (low-rank or masked) and sketched updates (quantized, rotated, subsampled) reduce uplink communication in federated learning by up to two orders of magnitude on convolutional and recurrent networks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Federated learning trains high-quality models on mobile devices while reducing uplink communication by up to 100 times.","strongest_claim":"Experiments on both convolutional and recurrent networks show that the proposed methods can reduce the communication cost by two orders of magnitude.","weakest_assumption":"That restricting updates to a low-rank or masked space or compressing them via quantization and subsampling preserves sufficient information for the aggregated global model to converge to high quality."}},"verdict_id":"5848767c-a797-431e-8e33-1dee5d14ec48"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:127a4df53ba0021b7d84c237b4cf05ff6218e4e60062ad6956bbee8764f37494","target":"record","created_at":"2026-07-04T22:18:31Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"3e7f111c4100d8648a1201de071b7bd64fce5e7e5e8239aff271f4c81380deab","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2016-10-18T09:11:51Z","title_canon_sha256":"e5b84b937bf0a2a632f1918628df4d3bcd260628291c3c43ae959b77d377f37d"},"schema_version":"1.0","source":{"id":"1610.05492","kind":"arxiv","version":2}},"canonical_sha256":"ef68e63b874de0ed80866262cced25e6096ba6b72bbc20af874be34d84f1077f","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"ef68e63b874de0ed80866262cced25e6096ba6b72bbc20af874be34d84f1077f","first_computed_at":"2026-07-04T22:18:31.925547Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-04T22:18:31.925547Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"2gYtBRspNBHdOnDVOi776SmevSX/ciNeyLK+iMThUeLqQ45DPGpQzCl7skoM54qPXATo/FLNArcbefVdOX1FBA==","signature_status":"signed_v1","signed_at":"2026-07-04T22:18:31.926054Z","signed_message":"canonical_sha256_bytes"},"source_id":"1610.05492","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:127a4df53ba0021b7d84c237b4cf05ff6218e4e60062ad6956bbee8764f37494","sha256:b63e08f7576fe093d1ef0198c684ca8f1e0af09d8beffe37d4407159d78c0829"],"state_sha256":"c6a851a24f3720f62115b3f97710246833de09cafb3b809b674945c784fc3be6"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"dr2pPqxbOy3A7s+EJ+zvbQp+KyfsdLk/O7EQy2Q2cSxLNtLlWuFlqZIBGZm6UWYj7fBcbxKdiq/WKNqy1gX8CQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-19T09:42:47.521606Z","bundle_sha256":"5954c271fbd6cbfb934dfa55d6a252cfae9c4a2f66a1df3db87ad1f7e9865d62"}}