{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:VNVRFFUB6E4GCHBL4C6O47LOOC","short_pith_number":"pith:VNVRFFUB","schema_version":"1.0","canonical_sha256":"ab6b129681f138611c2be0bcee7d6e70a32e5d4066fa6a728e5723ae9f886dd3","source":{"kind":"arxiv","id":"1810.07052","version":3},"attestation_state":"computed","paper":{"title":"Shallow-Deep Networks: Understanding and Mitigating Network Overthinking","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Sanghyun Hong, Tudor Dumitras, Yigitcan Kaya","submitted_at":"2018-10-16T14:51:13Z","abstract_excerpt":"We characterize a prevalent weakness of deep neural networks (DNNs)---overthinking---which occurs when a DNN can reach correct predictions before its final layer. Overthinking is computationally wasteful, and it can also be destructive when, by the final layer, a correct prediction changes into a misclassification. Understanding overthinking requires studying how each prediction evolves during a DNN's forward pass, which conventionally is opaque. For prediction transparency, we propose the Shallow-Deep Network (SDN), a generic modification to off-the-shelf DNNs that introduces internal classif"},"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":"1810.07052","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2018-10-16T14:51:13Z","cross_cats_sorted":["cs.CV","stat.ML"],"title_canon_sha256":"921253429aaca540ea1736e1f92c2ed0080466a608b234a4b15b723fc6a01148","abstract_canon_sha256":"a59dc6c7f080a9d15c7318c111cb1ec3ca05d6d499fe61b5e4de733cec2cd637"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:46:41.532441Z","signature_b64":"caRplcztkRBu2ac4lNAFhfroQFRe+c02w3MrxzRyH4qW04vzFpwbF3wh4qlZh4Q/B+shOQv9Te2+00i9iR6sCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ab6b129681f138611c2be0bcee7d6e70a32e5d4066fa6a728e5723ae9f886dd3","last_reissued_at":"2026-05-17T23:46:41.531787Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:46:41.531787Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Shallow-Deep Networks: Understanding and Mitigating Network Overthinking","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Sanghyun Hong, Tudor Dumitras, Yigitcan Kaya","submitted_at":"2018-10-16T14:51:13Z","abstract_excerpt":"We characterize a prevalent weakness of deep neural networks (DNNs)---overthinking---which occurs when a DNN can reach correct predictions before its final layer. Overthinking is computationally wasteful, and it can also be destructive when, by the final layer, a correct prediction changes into a misclassification. Understanding overthinking requires studying how each prediction evolves during a DNN's forward pass, which conventionally is opaque. For prediction transparency, we propose the Shallow-Deep Network (SDN), a generic modification to off-the-shelf DNNs that introduces internal classif"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1810.07052","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":""},"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":"1810.07052","created_at":"2026-05-17T23:46:41.531885+00:00"},{"alias_kind":"arxiv_version","alias_value":"1810.07052v3","created_at":"2026-05-17T23:46:41.531885+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1810.07052","created_at":"2026-05-17T23:46:41.531885+00:00"},{"alias_kind":"pith_short_12","alias_value":"VNVRFFUB6E4G","created_at":"2026-05-18T12:32:59.047623+00:00"},{"alias_kind":"pith_short_16","alias_value":"VNVRFFUB6E4GCHBL","created_at":"2026-05-18T12:32:59.047623+00:00"},{"alias_kind":"pith_short_8","alias_value":"VNVRFFUB","created_at":"2026-05-18T12:32:59.047623+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.05527","citing_title":"EdgeServing: Deadline-Aware Multi-DNN Serving at the Edge","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14789","citing_title":"A Comparative Study of CNN Optimization Methods for Edge AI: Exploring the Role of Early Exits","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VNVRFFUB6E4GCHBL4C6O47LOOC","json":"https://pith.science/pith/VNVRFFUB6E4GCHBL4C6O47LOOC.json","graph_json":"https://pith.science/api/pith-number/VNVRFFUB6E4GCHBL4C6O47LOOC/graph.json","events_json":"https://pith.science/api/pith-number/VNVRFFUB6E4GCHBL4C6O47LOOC/events.json","paper":"https://pith.science/paper/VNVRFFUB"},"agent_actions":{"view_html":"https://pith.science/pith/VNVRFFUB6E4GCHBL4C6O47LOOC","download_json":"https://pith.science/pith/VNVRFFUB6E4GCHBL4C6O47LOOC.json","view_paper":"https://pith.science/paper/VNVRFFUB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1810.07052&json=true","fetch_graph":"https://pith.science/api/pith-number/VNVRFFUB6E4GCHBL4C6O47LOOC/graph.json","fetch_events":"https://pith.science/api/pith-number/VNVRFFUB6E4GCHBL4C6O47LOOC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VNVRFFUB6E4GCHBL4C6O47LOOC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VNVRFFUB6E4GCHBL4C6O47LOOC/action/storage_attestation","attest_author":"https://pith.science/pith/VNVRFFUB6E4GCHBL4C6O47LOOC/action/author_attestation","sign_citation":"https://pith.science/pith/VNVRFFUB6E4GCHBL4C6O47LOOC/action/citation_signature","submit_replication":"https://pith.science/pith/VNVRFFUB6E4GCHBL4C6O47LOOC/action/replication_record"}},"created_at":"2026-05-17T23:46:41.531885+00:00","updated_at":"2026-05-17T23:46:41.531885+00:00"}