{"paper":{"title":"CAWI: Copula-Aligned Weight Initialization for Randomized Neural Networks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"CAWI samples randomized neural network weights from data-fitted copulas to capture inter-feature dependence and raise accuracy.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Mohd. Arshad, M. Tanveer, Mushir Akhtar","submitted_at":"2026-05-12T15:46:23Z","abstract_excerpt":"Randomized neural networks (RdNNs) enable efficient, backpropagation-free training by freezing randomly initialized input-to-hidden weights, which permits a closed-form solution for the output layer. However, conventional random initialization is blind to inter-feature dependence, ignoring correlations, asymmetries, and tail dependence in the data, which degrades conditioning and predictive performance. To the best of our knowledge, this limitation remains unaddressed in the RdNN literature. To close this gap, we propose CAWI (Copula-Aligned Weight Initialization), a framework that draws input"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"CAWI consistently delivers significant improvements in predictive performance over conventional random initialization across 83 diverse classification benchmarks and two biomedical datasets using standard shallow and deep RdNN architectures.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That sampling the input-to-hidden weights from a fitted copula will improve the conditioning and predictive performance of the closed-form output-layer solution without introducing new instabilities or requiring changes to the solver.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CAWI replaces standard random initialization of input-to-hidden weights in randomized neural networks with samples drawn from a data-fitted copula that preserves observed feature dependencies, yielding consistent accuracy gains on 83 classification benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"CAWI samples randomized neural network weights from data-fitted copulas to capture inter-feature dependence and raise accuracy.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"b021bb2861eda90549ba861fd31a4321ba6136b2e4a05e7b6f27c555312ce69c"},"source":{"id":"2605.12580","kind":"arxiv","version":1},"verdict":{"id":"2a61142c-3ccb-4607-9d85-600628da197c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T21:59:54.957462Z","strongest_claim":"CAWI consistently delivers significant improvements in predictive performance over conventional random initialization across 83 diverse classification benchmarks and two biomedical datasets using standard shallow and deep RdNN architectures.","one_line_summary":"CAWI replaces standard random initialization of input-to-hidden weights in randomized neural networks with samples drawn from a data-fitted copula that preserves observed feature dependencies, yielding consistent accuracy gains on 83 classification benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That sampling the input-to-hidden weights from a fitted copula will improve the conditioning and predictive performance of the closed-form output-layer solution without introducing new instabilities or requiring changes to the solver.","pith_extraction_headline":"CAWI samples randomized neural network weights from data-fitted copulas to capture inter-feature dependence and raise accuracy."},"references":{"count":105,"sample":[{"doi":"","year":2000,"title":"Langley , title =","work_id":"6cd283dc-0548-45e9-af07-6bc1005593ad","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1980,"title":"T. M. Mitchell. The Need for Biases in Learning Generalizations. 1980","work_id":"6b09bca6-ef5d-4a83-8c8e-219f23cbd761","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"M. J. Kearns , title =","work_id":"8efd8073-6f5d-45c5-94d5-62d366b52518","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1983,"title":"Machine Learning: An Artificial Intelligence Approach, Vol. I. 1983","work_id":"51835800-f16e-4534-8339-d3ea09147556","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2000,"title":"R. O. Duda and P. E. Hart and D. G. Stork. Pattern Classification. 2000","work_id":"a24cb892-7ac9-4509-8f5b-abbdfb15998b","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":105,"snapshot_sha256":"8fdab9df4f0a8e8679de56f9ad872ce499115a49df0a9b36fcdbcf21a8fd11ae","internal_anchors":3},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f0cf294e533311e66226733096dc05c6dcf32dc8d261671f070ce95ff381f766"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}