{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2017:LX7Y7DU64XSHI2TKAPCHZRRCBV","short_pith_number":"pith:LX7Y7DU6","schema_version":"1.0","canonical_sha256":"5dff8f8e9ee5e4746a6a03c47cc6220d5870b426200f9ae307363beca814152a","source":{"kind":"arxiv","id":"1709.04057","version":2},"attestation_state":"computed","paper":{"title":"Parallelizing Linear Recurrent Neural Nets Over Sequence Length","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.NE","authors_text":"Chris Cundy, Eric Martin","submitted_at":"2017-09-12T20:52:22Z","abstract_excerpt":"Recurrent neural networks (RNNs) are widely used to model sequential data but their non-linear dependencies between sequence elements prevent parallelizing training over sequence length. We show the training of RNNs with only linear sequential dependencies can be parallelized over the sequence length using the parallel scan algorithm, leading to rapid training on long sequences even with small minibatch size. We develop a parallel linear recurrence CUDA kernel and show that it can be applied to immediately speed up training and inference of several state of the art RNN architectures by up to 9"},"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":"1709.04057","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.NE","submitted_at":"2017-09-12T20:52:22Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"f9f737f2913dcbb84496927450cfcf966202324235725150b202ab396acb5f66","abstract_canon_sha256":"67974c1357b735012b02b21c0782409214d9091382c4121f81342b726817d552"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:22:49.265854Z","signature_b64":"YjsQZ0GuGxDdLQFqgUZi8Pq1TW8R03mGgFK51/0LhL1lGtJSRwoppRKqdCekd1SSo8Qbt4RcTyxUp0dj04jnAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5dff8f8e9ee5e4746a6a03c47cc6220d5870b426200f9ae307363beca814152a","last_reissued_at":"2026-05-18T00:22:49.265356Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:22:49.265356Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Parallelizing Linear Recurrent Neural Nets Over Sequence Length","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.NE","authors_text":"Chris Cundy, Eric Martin","submitted_at":"2017-09-12T20:52:22Z","abstract_excerpt":"Recurrent neural networks (RNNs) are widely used to model sequential data but their non-linear dependencies between sequence elements prevent parallelizing training over sequence length. We show the training of RNNs with only linear sequential dependencies can be parallelized over the sequence length using the parallel scan algorithm, leading to rapid training on long sequences even with small minibatch size. We develop a parallel linear recurrence CUDA kernel and show that it can be applied to immediately speed up training and inference of several state of the art RNN architectures by up to 9"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1709.04057","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":""},"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":"1709.04057","created_at":"2026-05-18T00:22:49.265434+00:00"},{"alias_kind":"arxiv_version","alias_value":"1709.04057v2","created_at":"2026-05-18T00:22:49.265434+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1709.04057","created_at":"2026-05-18T00:22:49.265434+00:00"},{"alias_kind":"pith_short_12","alias_value":"LX7Y7DU64XSH","created_at":"2026-05-18T12:31:28.150371+00:00"},{"alias_kind":"pith_short_16","alias_value":"LX7Y7DU64XSHI2TK","created_at":"2026-05-18T12:31:28.150371+00:00"},{"alias_kind":"pith_short_8","alias_value":"LX7Y7DU6","created_at":"2026-05-18T12:31:28.150371+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":5,"sample":[{"citing_arxiv_id":"2504.17333","citing_title":"Fine-Grained Fusion: The Missing Piece in Area-Efficient State Space Model Acceleration","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2605.15216","citing_title":"Hardware-Software Co-Design of Scalable, Energy-Efficient Analog Recurrent Computations","ref_index":61,"is_internal_anchor":true},{"citing_arxiv_id":"2602.14814","citing_title":"Learning State-Tracking from Code Using Linear RNNs","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2402.19427","citing_title":"Griffin: Mixing Gated Linear Recurrences with Local Attention for Efficient Language Models","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13807","citing_title":"Parallel Scan Recurrent Neural Quantum States for Scalable Variational Monte Carlo","ref_index":25,"is_internal_anchor":true},{"citing_arxiv_id":"2305.13048","citing_title":"RWKV: Reinventing RNNs for the Transformer Era","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LX7Y7DU64XSHI2TKAPCHZRRCBV","json":"https://pith.science/pith/LX7Y7DU64XSHI2TKAPCHZRRCBV.json","graph_json":"https://pith.science/api/pith-number/LX7Y7DU64XSHI2TKAPCHZRRCBV/graph.json","events_json":"https://pith.science/api/pith-number/LX7Y7DU64XSHI2TKAPCHZRRCBV/events.json","paper":"https://pith.science/paper/LX7Y7DU6"},"agent_actions":{"view_html":"https://pith.science/pith/LX7Y7DU64XSHI2TKAPCHZRRCBV","download_json":"https://pith.science/pith/LX7Y7DU64XSHI2TKAPCHZRRCBV.json","view_paper":"https://pith.science/paper/LX7Y7DU6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1709.04057&json=true","fetch_graph":"https://pith.science/api/pith-number/LX7Y7DU64XSHI2TKAPCHZRRCBV/graph.json","fetch_events":"https://pith.science/api/pith-number/LX7Y7DU64XSHI2TKAPCHZRRCBV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LX7Y7DU64XSHI2TKAPCHZRRCBV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LX7Y7DU64XSHI2TKAPCHZRRCBV/action/storage_attestation","attest_author":"https://pith.science/pith/LX7Y7DU64XSHI2TKAPCHZRRCBV/action/author_attestation","sign_citation":"https://pith.science/pith/LX7Y7DU64XSHI2TKAPCHZRRCBV/action/citation_signature","submit_replication":"https://pith.science/pith/LX7Y7DU64XSHI2TKAPCHZRRCBV/action/replication_record"}},"created_at":"2026-05-18T00:22:49.265434+00:00","updated_at":"2026-05-18T00:22:49.265434+00:00"}