{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:JWFTTWK6ZTJFNCVRJ53TD6YGFH","short_pith_number":"pith:JWFTTWK6","schema_version":"1.0","canonical_sha256":"4d8b39d95eccd2568ab14f7731fb0629c94de309ea3c16b92ef6a37d325c4967","source":{"kind":"arxiv","id":"1805.09208","version":2},"attestation_state":"computed","paper":{"title":"Pushing the bounds of dropout","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"stat.ML","authors_text":"Charles Blundell, Chris Dyer, G\\'abor Melis, Karl Moritz Hermann, Phil Blunsom, Tom\\'a\\v{s} Ko\\v{c}isk\\'y","submitted_at":"2018-05-23T14:55:39Z","abstract_excerpt":"We show that dropout training is best understood as performing MAP estimation concurrently for a family of conditional models whose objectives are themselves lower bounded by the original dropout objective. This discovery allows us to pick any model from this family after training, which leads to a substantial improvement on regularisation-heavy language modelling. The family includes models that compute a power mean over the sampled dropout masks, and their less stochastic subvariants with tighter and higher lower bounds than the fully stochastic dropout objective. We argue that since the det"},"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":"1805.09208","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2018-05-23T14:55:39Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"3c1e2748071e066e740d395696b7c6906ccde3c650e97680b725b51a447ffd19","abstract_canon_sha256":"d2e3a271195cabbf6ae7f96d124ef104b165daa6332e2856984291723a71fbf9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:04:39.343276Z","signature_b64":"+MbmNm2sLHOH9SSgz3C7ixlt5l6cYa2gTHQs8x6k+5AtIKmZ3SHzt0rQzqKaOr+XiqC9Sl+b58JvMp1nqEPWDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4d8b39d95eccd2568ab14f7731fb0629c94de309ea3c16b92ef6a37d325c4967","last_reissued_at":"2026-05-18T00:04:39.342803Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:04:39.342803Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Pushing the bounds of dropout","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"stat.ML","authors_text":"Charles Blundell, Chris Dyer, G\\'abor Melis, Karl Moritz Hermann, Phil Blunsom, Tom\\'a\\v{s} Ko\\v{c}isk\\'y","submitted_at":"2018-05-23T14:55:39Z","abstract_excerpt":"We show that dropout training is best understood as performing MAP estimation concurrently for a family of conditional models whose objectives are themselves lower bounded by the original dropout objective. This discovery allows us to pick any model from this family after training, which leads to a substantial improvement on regularisation-heavy language modelling. The family includes models that compute a power mean over the sampled dropout masks, and their less stochastic subvariants with tighter and higher lower bounds than the fully stochastic dropout objective. We argue that since the det"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1805.09208","kind":"arxiv","version":2},"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":"1805.09208","created_at":"2026-05-18T00:04:39.342867+00:00"},{"alias_kind":"arxiv_version","alias_value":"1805.09208v2","created_at":"2026-05-18T00:04:39.342867+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1805.09208","created_at":"2026-05-18T00:04:39.342867+00:00"},{"alias_kind":"pith_short_12","alias_value":"JWFTTWK6ZTJF","created_at":"2026-05-18T12:32:33.847187+00:00"},{"alias_kind":"pith_short_16","alias_value":"JWFTTWK6ZTJFNCVR","created_at":"2026-05-18T12:32:33.847187+00:00"},{"alias_kind":"pith_short_8","alias_value":"JWFTTWK6","created_at":"2026-05-18T12:32:33.847187+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"1909.01792","citing_title":"Mogrifier LSTM","ref_index":35,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JWFTTWK6ZTJFNCVRJ53TD6YGFH","json":"https://pith.science/pith/JWFTTWK6ZTJFNCVRJ53TD6YGFH.json","graph_json":"https://pith.science/api/pith-number/JWFTTWK6ZTJFNCVRJ53TD6YGFH/graph.json","events_json":"https://pith.science/api/pith-number/JWFTTWK6ZTJFNCVRJ53TD6YGFH/events.json","paper":"https://pith.science/paper/JWFTTWK6"},"agent_actions":{"view_html":"https://pith.science/pith/JWFTTWK6ZTJFNCVRJ53TD6YGFH","download_json":"https://pith.science/pith/JWFTTWK6ZTJFNCVRJ53TD6YGFH.json","view_paper":"https://pith.science/paper/JWFTTWK6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1805.09208&json=true","fetch_graph":"https://pith.science/api/pith-number/JWFTTWK6ZTJFNCVRJ53TD6YGFH/graph.json","fetch_events":"https://pith.science/api/pith-number/JWFTTWK6ZTJFNCVRJ53TD6YGFH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JWFTTWK6ZTJFNCVRJ53TD6YGFH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JWFTTWK6ZTJFNCVRJ53TD6YGFH/action/storage_attestation","attest_author":"https://pith.science/pith/JWFTTWK6ZTJFNCVRJ53TD6YGFH/action/author_attestation","sign_citation":"https://pith.science/pith/JWFTTWK6ZTJFNCVRJ53TD6YGFH/action/citation_signature","submit_replication":"https://pith.science/pith/JWFTTWK6ZTJFNCVRJ53TD6YGFH/action/replication_record"}},"created_at":"2026-05-18T00:04:39.342867+00:00","updated_at":"2026-05-18T00:04:39.342867+00:00"}