{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:2LBHIH3ABFIODT6UVI5VF27D4E","short_pith_number":"pith:2LBHIH3A","schema_version":"1.0","canonical_sha256":"d2c2741f600950e1cfd4aa3b52ebe3e122df5fb9e4d45add45ea83d83f678408","source":{"kind":"arxiv","id":"2009.10685","version":3},"attestation_state":"computed","paper":{"title":"Tensor Programs III: Neural Matrix Laws","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.PR"],"primary_cat":"cs.NE","authors_text":"Greg Yang","submitted_at":"2020-09-22T16:59:24Z","abstract_excerpt":"In a neural network (NN), *weight matrices* linearly transform inputs into *preactivations* that are then transformed nonlinearly into *activations*. A typical NN interleaves multitudes of such linear and nonlinear transforms to express complex functions. Thus, the (pre-)activations depend on the weights in an intricate manner. We show that, surprisingly, (pre-)activations of a randomly initialized NN become *independent* from the weights as the NN's widths tend to infinity, in the sense of asymptotic freeness in random matrix theory. We call this the Free Independence Principle (FIP), which h"},"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":"2009.10685","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.NE","submitted_at":"2020-09-22T16:59:24Z","cross_cats_sorted":["math.PR"],"title_canon_sha256":"e81af3b70d3c1024f48ff6e6fb6c1d6491f87fb0ae679ead3cf446e3732099a3","abstract_canon_sha256":"b1ab4e8bed2826afb981f2206553db76d26c9f675b6dd6bcf51c99166ef6a47b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:38:30.349776Z","signature_b64":"KpjdDQnxEz/qz0h5FMKWw0tHZxUyCb3yrfJPxSD7SggxwvwT7c7vOklubwAlATpYI4Lsmn/AjY5109OsRn4nDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d2c2741f600950e1cfd4aa3b52ebe3e122df5fb9e4d45add45ea83d83f678408","last_reissued_at":"2026-07-05T02:38:30.349334Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:38:30.349334Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tensor Programs III: Neural Matrix Laws","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.PR"],"primary_cat":"cs.NE","authors_text":"Greg Yang","submitted_at":"2020-09-22T16:59:24Z","abstract_excerpt":"In a neural network (NN), *weight matrices* linearly transform inputs into *preactivations* that are then transformed nonlinearly into *activations*. A typical NN interleaves multitudes of such linear and nonlinear transforms to express complex functions. Thus, the (pre-)activations depend on the weights in an intricate manner. We show that, surprisingly, (pre-)activations of a randomly initialized NN become *independent* from the weights as the NN's widths tend to infinity, in the sense of asymptotic freeness in random matrix theory. We call this the Free Independence Principle (FIP), which h"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2009.10685","kind":"arxiv","version":3},"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/2009.10685/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":"2009.10685","created_at":"2026-07-05T02:38:30.349393+00:00"},{"alias_kind":"arxiv_version","alias_value":"2009.10685v3","created_at":"2026-07-05T02:38:30.349393+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2009.10685","created_at":"2026-07-05T02:38:30.349393+00:00"},{"alias_kind":"pith_short_12","alias_value":"2LBHIH3ABFIO","created_at":"2026-07-05T02:38:30.349393+00:00"},{"alias_kind":"pith_short_16","alias_value":"2LBHIH3ABFIODT6U","created_at":"2026-07-05T02:38:30.349393+00:00"},{"alias_kind":"pith_short_8","alias_value":"2LBHIH3A","created_at":"2026-07-05T02:38:30.349393+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.30860","citing_title":"Bayesian Inference with Shaped Deep Non-linear MLPs","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18180","citing_title":"Canonical Regularisation of Wide Feature-Learning Neural Networks","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2603.00541","citing_title":"Spectral Condition for $\\mu$P under Width-Depth Scaling","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05113","citing_title":"How Long Does Infinite Width Last? Signal Propagation in Long-Range Linear Recurrences","ref_index":43,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2LBHIH3ABFIODT6UVI5VF27D4E","json":"https://pith.science/pith/2LBHIH3ABFIODT6UVI5VF27D4E.json","graph_json":"https://pith.science/api/pith-number/2LBHIH3ABFIODT6UVI5VF27D4E/graph.json","events_json":"https://pith.science/api/pith-number/2LBHIH3ABFIODT6UVI5VF27D4E/events.json","paper":"https://pith.science/paper/2LBHIH3A"},"agent_actions":{"view_html":"https://pith.science/pith/2LBHIH3ABFIODT6UVI5VF27D4E","download_json":"https://pith.science/pith/2LBHIH3ABFIODT6UVI5VF27D4E.json","view_paper":"https://pith.science/paper/2LBHIH3A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2009.10685&json=true","fetch_graph":"https://pith.science/api/pith-number/2LBHIH3ABFIODT6UVI5VF27D4E/graph.json","fetch_events":"https://pith.science/api/pith-number/2LBHIH3ABFIODT6UVI5VF27D4E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2LBHIH3ABFIODT6UVI5VF27D4E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2LBHIH3ABFIODT6UVI5VF27D4E/action/storage_attestation","attest_author":"https://pith.science/pith/2LBHIH3ABFIODT6UVI5VF27D4E/action/author_attestation","sign_citation":"https://pith.science/pith/2LBHIH3ABFIODT6UVI5VF27D4E/action/citation_signature","submit_replication":"https://pith.science/pith/2LBHIH3ABFIODT6UVI5VF27D4E/action/replication_record"}},"created_at":"2026-07-05T02:38:30.349393+00:00","updated_at":"2026-07-05T02:38:30.349393+00:00"}