{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:EVO3RZ2ZDTZAOFDEY5YS6AWGBD","short_pith_number":"pith:EVO3RZ2Z","schema_version":"1.0","canonical_sha256":"255db8e7591cf2071464c7712f02c608caa9d8c2af67b1bd63916e700d7cdbe0","source":{"kind":"arxiv","id":"2108.12659","version":4},"attestation_state":"computed","paper":{"title":"DKM: Differentiable K-Means Clustering Layer for Neural Network Compression","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Keivan A. Vahid, Minsik Cho, Mohammad Rastegari, Saurabh Adya","submitted_at":"2021-08-28T14:35:41Z","abstract_excerpt":"Deep neural network (DNN) model compression for efficient on-device inference is becoming increasingly important to reduce memory requirements and keep user data on-device. To this end, we propose a novel differentiable k-means clustering layer (DKM) and its application to train-time weight clustering-based DNN model compression. DKM casts k-means clustering as an attention problem and enables joint optimization of the DNN parameters and clustering centroids. Unlike prior works that rely on additional regularizers and parameters, DKM-based compression keeps the original loss function and model"},"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":"2108.12659","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2021-08-28T14:35:41Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"89fa71cabb148a1ef446b5cbdf4f678460d0489be7837cdc0d0e509709809aa0","abstract_canon_sha256":"ae0f5df620daf2aa021c45f214b79f54e88057cf8facf9642dfa423e955599a0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:58:17.159003Z","signature_b64":"viq9GpLOxEuQDRZzVjwjw1tgEHo/Q1Jv9x8nkyC7tUY6wzeAjvTyEKCyJxKAWOuav0LfM+PJ5zzxEinr6JqlCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"255db8e7591cf2071464c7712f02c608caa9d8c2af67b1bd63916e700d7cdbe0","last_reissued_at":"2026-07-05T03:58:17.158555Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:58:17.158555Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DKM: Differentiable K-Means Clustering Layer for Neural Network Compression","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Keivan A. Vahid, Minsik Cho, Mohammad Rastegari, Saurabh Adya","submitted_at":"2021-08-28T14:35:41Z","abstract_excerpt":"Deep neural network (DNN) model compression for efficient on-device inference is becoming increasingly important to reduce memory requirements and keep user data on-device. To this end, we propose a novel differentiable k-means clustering layer (DKM) and its application to train-time weight clustering-based DNN model compression. DKM casts k-means clustering as an attention problem and enables joint optimization of the DNN parameters and clustering centroids. Unlike prior works that rely on additional regularizers and parameters, DKM-based compression keeps the original loss function and model"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2108.12659","kind":"arxiv","version":4},"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/2108.12659/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":"2108.12659","created_at":"2026-07-05T03:58:17.158607+00:00"},{"alias_kind":"arxiv_version","alias_value":"2108.12659v4","created_at":"2026-07-05T03:58:17.158607+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2108.12659","created_at":"2026-07-05T03:58:17.158607+00:00"},{"alias_kind":"pith_short_12","alias_value":"EVO3RZ2ZDTZA","created_at":"2026-07-05T03:58:17.158607+00:00"},{"alias_kind":"pith_short_16","alias_value":"EVO3RZ2ZDTZAOFDE","created_at":"2026-07-05T03:58:17.158607+00:00"},{"alias_kind":"pith_short_8","alias_value":"EVO3RZ2Z","created_at":"2026-07-05T03:58:17.158607+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.23190","citing_title":"RAT: RunAnyThing via Fully Automated Environment Configuration","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23172","citing_title":"Efficient VQ-QAT and Mixed Vector/Linear quantized Neural Networks","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EVO3RZ2ZDTZAOFDEY5YS6AWGBD","json":"https://pith.science/pith/EVO3RZ2ZDTZAOFDEY5YS6AWGBD.json","graph_json":"https://pith.science/api/pith-number/EVO3RZ2ZDTZAOFDEY5YS6AWGBD/graph.json","events_json":"https://pith.science/api/pith-number/EVO3RZ2ZDTZAOFDEY5YS6AWGBD/events.json","paper":"https://pith.science/paper/EVO3RZ2Z"},"agent_actions":{"view_html":"https://pith.science/pith/EVO3RZ2ZDTZAOFDEY5YS6AWGBD","download_json":"https://pith.science/pith/EVO3RZ2ZDTZAOFDEY5YS6AWGBD.json","view_paper":"https://pith.science/paper/EVO3RZ2Z","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2108.12659&json=true","fetch_graph":"https://pith.science/api/pith-number/EVO3RZ2ZDTZAOFDEY5YS6AWGBD/graph.json","fetch_events":"https://pith.science/api/pith-number/EVO3RZ2ZDTZAOFDEY5YS6AWGBD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EVO3RZ2ZDTZAOFDEY5YS6AWGBD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EVO3RZ2ZDTZAOFDEY5YS6AWGBD/action/storage_attestation","attest_author":"https://pith.science/pith/EVO3RZ2ZDTZAOFDEY5YS6AWGBD/action/author_attestation","sign_citation":"https://pith.science/pith/EVO3RZ2ZDTZAOFDEY5YS6AWGBD/action/citation_signature","submit_replication":"https://pith.science/pith/EVO3RZ2ZDTZAOFDEY5YS6AWGBD/action/replication_record"}},"created_at":"2026-07-05T03:58:17.158607+00:00","updated_at":"2026-07-05T03:58:17.158607+00:00"}