{"id":"a19e47b1-e501-4d3d-ad76-5afccce741d7","arxiv_id":"2504.11981","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A dot-product-based reservoir representation enables a fully digital delayed-feedback reservoir that matches deep learning accuracy on multivariate time-series classification with far smaller FPGA circuits.","lead":"This paper proposes a new way to turn reservoir-computer outputs into fixed-size features for time-series classification, using dot products of shifted reservoir states. It shows that this feature method lets a fully digital delayed-feedback reservoir run on a small FPGA with accuracy close to deep learning models but far fewer hardware resources.","discovery_kind":"new_method","skeptic_critique":null,"referee_report":null,"author_rebuttal":null,"desk_editor":null,"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-16T12:40:48.610150+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}