{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:KORTYP52XUKTDES2NF3U3CLO55","short_pith_number":"pith:KORTYP52","schema_version":"1.0","canonical_sha256":"53a33c3fbabd1531925a69774d896eef54c2a62a27dae50e79ced22f1ec7646b","source":{"kind":"arxiv","id":"2106.13739","version":1},"attestation_state":"computed","paper":{"title":"Re-parameterizing VAEs for stability","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"David Dehaene, R\\'emy Brossard","submitted_at":"2021-06-25T16:19:09Z","abstract_excerpt":"We propose a theoretical approach towards the training numerical stability of Variational AutoEncoders (VAE). Our work is motivated by recent studies empowering VAEs to reach state of the art generative results on complex image datasets. These very deep VAE architectures, as well as VAEs using more complex output distributions, highlight a tendency to haphazardly produce high training gradients as well as NaN losses. The empirical fixes proposed to train them despite their limitations are neither fully theoretically grounded nor generally sufficient in practice. Building on this, we localize t"},"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":"2106.13739","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-25T16:19:09Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"7e0e500950e0ff94f1661298a444dfe1bc764db723ca8924534c122888e0bc39","abstract_canon_sha256":"6ddf1b5749f4dc111e1770bcd29fe2592d2085ba77209af18d0d6d7a62d99207"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:52:25.657567Z","signature_b64":"teBVjHnuBrAyRf9Gs5xcjtPbJ95wC2supNLgvnc0Q6iZOm7xsyuvMkeJLovBYZUJPwAV9UWTjzV3tYzR5FDzDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"53a33c3fbabd1531925a69774d896eef54c2a62a27dae50e79ced22f1ec7646b","last_reissued_at":"2026-07-05T02:52:25.657141Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:52:25.657141Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Re-parameterizing VAEs for stability","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"David Dehaene, R\\'emy Brossard","submitted_at":"2021-06-25T16:19:09Z","abstract_excerpt":"We propose a theoretical approach towards the training numerical stability of Variational AutoEncoders (VAE). Our work is motivated by recent studies empowering VAEs to reach state of the art generative results on complex image datasets. These very deep VAE architectures, as well as VAEs using more complex output distributions, highlight a tendency to haphazardly produce high training gradients as well as NaN losses. The empirical fixes proposed to train them despite their limitations are neither fully theoretically grounded nor generally sufficient in practice. Building on this, we localize t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.13739","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/2106.13739/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":"2106.13739","created_at":"2026-07-05T02:52:25.657198+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.13739v1","created_at":"2026-07-05T02:52:25.657198+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.13739","created_at":"2026-07-05T02:52:25.657198+00:00"},{"alias_kind":"pith_short_12","alias_value":"KORTYP52XUKT","created_at":"2026-07-05T02:52:25.657198+00:00"},{"alias_kind":"pith_short_16","alias_value":"KORTYP52XUKTDES2","created_at":"2026-07-05T02:52:25.657198+00:00"},{"alias_kind":"pith_short_8","alias_value":"KORTYP52","created_at":"2026-07-05T02:52:25.657198+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2501.07763","citing_title":"On the Statistical Capacity of Deep Generative Models","ref_index":11,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KORTYP52XUKTDES2NF3U3CLO55","json":"https://pith.science/pith/KORTYP52XUKTDES2NF3U3CLO55.json","graph_json":"https://pith.science/api/pith-number/KORTYP52XUKTDES2NF3U3CLO55/graph.json","events_json":"https://pith.science/api/pith-number/KORTYP52XUKTDES2NF3U3CLO55/events.json","paper":"https://pith.science/paper/KORTYP52"},"agent_actions":{"view_html":"https://pith.science/pith/KORTYP52XUKTDES2NF3U3CLO55","download_json":"https://pith.science/pith/KORTYP52XUKTDES2NF3U3CLO55.json","view_paper":"https://pith.science/paper/KORTYP52","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.13739&json=true","fetch_graph":"https://pith.science/api/pith-number/KORTYP52XUKTDES2NF3U3CLO55/graph.json","fetch_events":"https://pith.science/api/pith-number/KORTYP52XUKTDES2NF3U3CLO55/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KORTYP52XUKTDES2NF3U3CLO55/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KORTYP52XUKTDES2NF3U3CLO55/action/storage_attestation","attest_author":"https://pith.science/pith/KORTYP52XUKTDES2NF3U3CLO55/action/author_attestation","sign_citation":"https://pith.science/pith/KORTYP52XUKTDES2NF3U3CLO55/action/citation_signature","submit_replication":"https://pith.science/pith/KORTYP52XUKTDES2NF3U3CLO55/action/replication_record"}},"created_at":"2026-07-05T02:52:25.657198+00:00","updated_at":"2026-07-05T02:52:25.657198+00:00"}