{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:6EEL43ISITE5XTMOPWDL2M5QOE","short_pith_number":"pith:6EEL43IS","schema_version":"1.0","canonical_sha256":"f108be6d1244c9dbcd8e7d86bd33b071291834fa40e2d4cd579d04d4f0b612c7","source":{"kind":"arxiv","id":"2306.01843","version":5},"attestation_state":"computed","paper":{"title":"Lifting Architectural Constraints of Injective Flows","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Armand Rousselot, Felix Draxler, Lea Zimmermann, Peter Sorrenson, Sander Hummerich, Ullrich K\\\"othe","submitted_at":"2023-06-02T18:03:03Z","abstract_excerpt":"Normalizing Flows explicitly maximize a full-dimensional likelihood on the training data. However, real data is typically only supported on a lower-dimensional manifold leading the model to expend significant compute on modeling noise. Injective Flows fix this by jointly learning a manifold and the distribution on it. So far, they have been limited by restrictive architectures and/or high computational cost. We lift both constraints by a new efficient estimator for the maximum likelihood loss, compatible with free-form bottleneck architectures. We further show that naively learning both the da"},"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":"2306.01843","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-02T18:03:03Z","cross_cats_sorted":[],"title_canon_sha256":"20fbea512d2e619fd0dd5e7f41b630aae35fcf941a406eaf257a43aeef514b97","abstract_canon_sha256":"a9b74155a82a3e1489fe051a90a8d2b456a9a75302e5517cc8efe45c9b395b95"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:37:13.433082Z","signature_b64":"8B/i6vw0vvvTEg+qApvHncEiPKXEMTdjSwCczSsRykmzYcNMIpzx3tpomV0iUJq7Hlb6F3HJ0tDCgfQpqJ/WAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f108be6d1244c9dbcd8e7d86bd33b071291834fa40e2d4cd579d04d4f0b612c7","last_reissued_at":"2026-07-05T08:37:13.432647Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:37:13.432647Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Lifting Architectural Constraints of Injective Flows","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Armand Rousselot, Felix Draxler, Lea Zimmermann, Peter Sorrenson, Sander Hummerich, Ullrich K\\\"othe","submitted_at":"2023-06-02T18:03:03Z","abstract_excerpt":"Normalizing Flows explicitly maximize a full-dimensional likelihood on the training data. However, real data is typically only supported on a lower-dimensional manifold leading the model to expend significant compute on modeling noise. Injective Flows fix this by jointly learning a manifold and the distribution on it. So far, they have been limited by restrictive architectures and/or high computational cost. We lift both constraints by a new efficient estimator for the maximum likelihood loss, compatible with free-form bottleneck architectures. We further show that naively learning both the da"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.01843","kind":"arxiv","version":5},"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/2306.01843/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":"2306.01843","created_at":"2026-07-05T08:37:13.432702+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.01843v5","created_at":"2026-07-05T08:37:13.432702+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.01843","created_at":"2026-07-05T08:37:13.432702+00:00"},{"alias_kind":"pith_short_12","alias_value":"6EEL43ISITE5","created_at":"2026-07-05T08:37:13.432702+00:00"},{"alias_kind":"pith_short_16","alias_value":"6EEL43ISITE5XTMO","created_at":"2026-07-05T08:37:13.432702+00:00"},{"alias_kind":"pith_short_8","alias_value":"6EEL43IS","created_at":"2026-07-05T08:37:13.432702+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.03517","citing_title":"Understanding Self-Supervised Learning via Latent Distribution Matching","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03517","citing_title":"Understanding Self-Supervised Learning via Latent Distribution Matching","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03517","citing_title":"Understanding Self-Supervised Learning via Latent Distribution Matching","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6EEL43ISITE5XTMOPWDL2M5QOE","json":"https://pith.science/pith/6EEL43ISITE5XTMOPWDL2M5QOE.json","graph_json":"https://pith.science/api/pith-number/6EEL43ISITE5XTMOPWDL2M5QOE/graph.json","events_json":"https://pith.science/api/pith-number/6EEL43ISITE5XTMOPWDL2M5QOE/events.json","paper":"https://pith.science/paper/6EEL43IS"},"agent_actions":{"view_html":"https://pith.science/pith/6EEL43ISITE5XTMOPWDL2M5QOE","download_json":"https://pith.science/pith/6EEL43ISITE5XTMOPWDL2M5QOE.json","view_paper":"https://pith.science/paper/6EEL43IS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.01843&json=true","fetch_graph":"https://pith.science/api/pith-number/6EEL43ISITE5XTMOPWDL2M5QOE/graph.json","fetch_events":"https://pith.science/api/pith-number/6EEL43ISITE5XTMOPWDL2M5QOE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6EEL43ISITE5XTMOPWDL2M5QOE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6EEL43ISITE5XTMOPWDL2M5QOE/action/storage_attestation","attest_author":"https://pith.science/pith/6EEL43ISITE5XTMOPWDL2M5QOE/action/author_attestation","sign_citation":"https://pith.science/pith/6EEL43ISITE5XTMOPWDL2M5QOE/action/citation_signature","submit_replication":"https://pith.science/pith/6EEL43ISITE5XTMOPWDL2M5QOE/action/replication_record"}},"created_at":"2026-07-05T08:37:13.432702+00:00","updated_at":"2026-07-05T08:37:13.432702+00:00"}