{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:OCV424MJ6YKPAX7ZTLRT36RRDV","short_pith_number":"pith:OCV424MJ","schema_version":"1.0","canonical_sha256":"70abcd7189f614f05ff99ae33dfa311d792fe39b07f083ae2b714d357bb2024a","source":{"kind":"arxiv","id":"2404.08602","version":2},"attestation_state":"computed","paper":{"title":"Sliding down the stairs: how correlated latent variables accelerate learning with neural networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cond-mat.stat-mech","cs.LG","math.PR","math.ST","stat.TH"],"primary_cat":"stat.ML","authors_text":"Lorenzo Bardone, Sebastian Goldt","submitted_at":"2024-04-12T17:01:25Z","abstract_excerpt":"Neural networks extract features from data using stochastic gradient descent (SGD). In particular, higher-order input cumulants (HOCs) are crucial for their performance. However, extracting information from the $p$th cumulant of $d$-dimensional inputs is computationally hard: the number of samples required to recover a single direction from an order-$p$ tensor (tensor PCA) using online SGD grows as $d^{p-1}$, which is prohibitive for high-dimensional inputs. This result raises the question of how neural networks extract relevant directions from the HOCs of their inputs efficiently. Here, we sh"},"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":"2404.08602","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2024-04-12T17:01:25Z","cross_cats_sorted":["cond-mat.stat-mech","cs.LG","math.PR","math.ST","stat.TH"],"title_canon_sha256":"1acb967c2cb31b79649ad1a2f6f14f667e150a58759159be6858c50a21d623bf","abstract_canon_sha256":"d2ecaee5ff61e4374b10042df9344eaa0df2bab6134b8122c6a23f7d3b7794b1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:20:42.227155Z","signature_b64":"OoamsvSid1X3JJ9Do/9P1KqDbs+DDLrlpXw54kB9FWZ7wnSY6CR7kYXu2Bdw1yCfm2XeOfoeEF6u+hZFv97UCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"70abcd7189f614f05ff99ae33dfa311d792fe39b07f083ae2b714d357bb2024a","last_reissued_at":"2026-07-05T09:20:42.226736Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:20:42.226736Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sliding down the stairs: how correlated latent variables accelerate learning with neural networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cond-mat.stat-mech","cs.LG","math.PR","math.ST","stat.TH"],"primary_cat":"stat.ML","authors_text":"Lorenzo Bardone, Sebastian Goldt","submitted_at":"2024-04-12T17:01:25Z","abstract_excerpt":"Neural networks extract features from data using stochastic gradient descent (SGD). In particular, higher-order input cumulants (HOCs) are crucial for their performance. However, extracting information from the $p$th cumulant of $d$-dimensional inputs is computationally hard: the number of samples required to recover a single direction from an order-$p$ tensor (tensor PCA) using online SGD grows as $d^{p-1}$, which is prohibitive for high-dimensional inputs. This result raises the question of how neural networks extract relevant directions from the HOCs of their inputs efficiently. Here, we sh"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.08602","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2404.08602/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":"2404.08602","created_at":"2026-07-05T09:20:42.226795+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.08602v2","created_at":"2026-07-05T09:20:42.226795+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.08602","created_at":"2026-07-05T09:20:42.226795+00:00"},{"alias_kind":"pith_short_12","alias_value":"OCV424MJ6YKP","created_at":"2026-07-05T09:20:42.226795+00:00"},{"alias_kind":"pith_short_16","alias_value":"OCV424MJ6YKPAX7Z","created_at":"2026-07-05T09:20:42.226795+00:00"},{"alias_kind":"pith_short_8","alias_value":"OCV424MJ","created_at":"2026-07-05T09:20:42.226795+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.07070","citing_title":"Scaling Laws and Representation Learning in Simple Hierarchical Languages: Transformers vs. Convolutional Architectures","ref_index":41,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OCV424MJ6YKPAX7ZTLRT36RRDV","json":"https://pith.science/pith/OCV424MJ6YKPAX7ZTLRT36RRDV.json","graph_json":"https://pith.science/api/pith-number/OCV424MJ6YKPAX7ZTLRT36RRDV/graph.json","events_json":"https://pith.science/api/pith-number/OCV424MJ6YKPAX7ZTLRT36RRDV/events.json","paper":"https://pith.science/paper/OCV424MJ"},"agent_actions":{"view_html":"https://pith.science/pith/OCV424MJ6YKPAX7ZTLRT36RRDV","download_json":"https://pith.science/pith/OCV424MJ6YKPAX7ZTLRT36RRDV.json","view_paper":"https://pith.science/paper/OCV424MJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.08602&json=true","fetch_graph":"https://pith.science/api/pith-number/OCV424MJ6YKPAX7ZTLRT36RRDV/graph.json","fetch_events":"https://pith.science/api/pith-number/OCV424MJ6YKPAX7ZTLRT36RRDV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OCV424MJ6YKPAX7ZTLRT36RRDV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OCV424MJ6YKPAX7ZTLRT36RRDV/action/storage_attestation","attest_author":"https://pith.science/pith/OCV424MJ6YKPAX7ZTLRT36RRDV/action/author_attestation","sign_citation":"https://pith.science/pith/OCV424MJ6YKPAX7ZTLRT36RRDV/action/citation_signature","submit_replication":"https://pith.science/pith/OCV424MJ6YKPAX7ZTLRT36RRDV/action/replication_record"}},"created_at":"2026-07-05T09:20:42.226795+00:00","updated_at":"2026-07-05T09:20:42.226795+00:00"}