{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:O2HCSKFHKU5AIKCMVMBS6FEJF7","short_pith_number":"pith:O2HCSKFH","schema_version":"1.0","canonical_sha256":"768e2928a7553a04284cab032f14892fe9726e8e445f62a2663440ac1110e4ff","source":{"kind":"arxiv","id":"1909.00719","version":4},"attestation_state":"computed","paper":{"title":"On the Expressiveness of Approximate Inference in Bayesian Neural Networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Andrew Y. K. Foong, David R. Burt, Richard E. Turner, Yingzhen Li","submitted_at":"2019-09-02T13:54:39Z","abstract_excerpt":"While Bayesian neural networks (BNNs) hold the promise of being flexible, well-calibrated statistical models, inference often requires approximations whose consequences are poorly understood. We study the quality of common variational methods in approximating the Bayesian predictive distribution. For single-hidden layer ReLU BNNs, we prove a fundamental limitation in function-space of two of the most commonly used distributions defined in weight-space: mean-field Gaussian and Monte Carlo dropout. We find there are simple cases where neither method can have substantially increased uncertainty i"},"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":"1909.00719","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2019-09-02T13:54:39Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"65cd9a6b89f0e7cb236ab63be88b646f2e838b24d6bb7ead29b351052ca8ed86","abstract_canon_sha256":"56e95cd9c0ac8eeedef41639bc1dcb52526202e9b29181f05e5c62befbcea0fb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:45:24.097641Z","signature_b64":"QVkDX0b5FsA8wbgmxGvgoXXTGYNGfZMGu3FfJU5UtZvchK63mVYW9ZFwdcFtAQhzr1gkkHddcuwftiRF7bmfDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"768e2928a7553a04284cab032f14892fe9726e8e445f62a2663440ac1110e4ff","last_reissued_at":"2026-07-05T01:45:24.097204Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:45:24.097204Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the Expressiveness of Approximate Inference in Bayesian Neural Networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Andrew Y. K. Foong, David R. Burt, Richard E. Turner, Yingzhen Li","submitted_at":"2019-09-02T13:54:39Z","abstract_excerpt":"While Bayesian neural networks (BNNs) hold the promise of being flexible, well-calibrated statistical models, inference often requires approximations whose consequences are poorly understood. We study the quality of common variational methods in approximating the Bayesian predictive distribution. For single-hidden layer ReLU BNNs, we prove a fundamental limitation in function-space of two of the most commonly used distributions defined in weight-space: mean-field Gaussian and Monte Carlo dropout. We find there are simple cases where neither method can have substantially increased uncertainty i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1909.00719","kind":"arxiv","version":4},"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/1909.00719/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":"1909.00719","created_at":"2026-07-05T01:45:24.097263+00:00"},{"alias_kind":"arxiv_version","alias_value":"1909.00719v4","created_at":"2026-07-05T01:45:24.097263+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1909.00719","created_at":"2026-07-05T01:45:24.097263+00:00"},{"alias_kind":"pith_short_12","alias_value":"O2HCSKFHKU5A","created_at":"2026-07-05T01:45:24.097263+00:00"},{"alias_kind":"pith_short_16","alias_value":"O2HCSKFHKU5AIKCM","created_at":"2026-07-05T01:45:24.097263+00:00"},{"alias_kind":"pith_short_8","alias_value":"O2HCSKFH","created_at":"2026-07-05T01:45:24.097263+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.12536","citing_title":"Information as Maximum-Caliber Deviation: A bridge between Integrated Information Theory and the Free Energy Principle","ref_index":158,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/O2HCSKFHKU5AIKCMVMBS6FEJF7","json":"https://pith.science/pith/O2HCSKFHKU5AIKCMVMBS6FEJF7.json","graph_json":"https://pith.science/api/pith-number/O2HCSKFHKU5AIKCMVMBS6FEJF7/graph.json","events_json":"https://pith.science/api/pith-number/O2HCSKFHKU5AIKCMVMBS6FEJF7/events.json","paper":"https://pith.science/paper/O2HCSKFH"},"agent_actions":{"view_html":"https://pith.science/pith/O2HCSKFHKU5AIKCMVMBS6FEJF7","download_json":"https://pith.science/pith/O2HCSKFHKU5AIKCMVMBS6FEJF7.json","view_paper":"https://pith.science/paper/O2HCSKFH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1909.00719&json=true","fetch_graph":"https://pith.science/api/pith-number/O2HCSKFHKU5AIKCMVMBS6FEJF7/graph.json","fetch_events":"https://pith.science/api/pith-number/O2HCSKFHKU5AIKCMVMBS6FEJF7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/O2HCSKFHKU5AIKCMVMBS6FEJF7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/O2HCSKFHKU5AIKCMVMBS6FEJF7/action/storage_attestation","attest_author":"https://pith.science/pith/O2HCSKFHKU5AIKCMVMBS6FEJF7/action/author_attestation","sign_citation":"https://pith.science/pith/O2HCSKFHKU5AIKCMVMBS6FEJF7/action/citation_signature","submit_replication":"https://pith.science/pith/O2HCSKFHKU5AIKCMVMBS6FEJF7/action/replication_record"}},"created_at":"2026-07-05T01:45:24.097263+00:00","updated_at":"2026-07-05T01:45:24.097263+00:00"}