[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-smooth-basis-models-regain-footing-in-tabular-regression":10},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":22,"tags":24,"sources":28,"feedback":32,"feedback_at":22,"cost_usd":32,"total_tokens":32},1357,"smooth-basis-models-regain-footing-in-tabular-regression","Smooth-basis models regain footing in tabular regression","Chebyshev polynomial and anisotropic RBF regressors match tree ensembles on CPU while narrowing generalisation gaps, per a new arXiv benchmark.","Smooth-basis regressors are back in the tabular game.\n\nThe authors built three scikit‑learn compatible models—a ridge‑regularised Chebyshev polynomial regressor, an anisotropic RBF network with data‑driven centre placement, and a hybrid Chebyshev model tree. They tested them on 55 regression datasets spanning multiple domains and compared results to tree ensembles, a pre‑trained transformer, and standard baselines. The transformer led on raw accuracy but required a GPU and suffered latency, limiting its use in typical CPU‑only environments. Among CPU‑friendly options, the smooth‑basis models and tree ensembles were statistically tied for accuracy, with the smooth models showing smaller gaps between training and test performance.\n\nThis matters because many applied science and industry pipelines run on modest hardware and value predictable, gradually varying predictions for downstream optimisation or sensitivity analysis. A tighter generalisation gap translates to more reliable surrogates without sacrificing speed.\n\nIn short, smooth‑basis methods deserve a seat at the regression table alongside trees, especially when deployment constraints rule out heavy transformers.","[\"regression\",\"ml-models\",\"tabular-data\"]","2026-06-16T04:00:00.000Z","2026-06-17T05:39:29.989Z","2026-06-17T05:39:32.825Z","published",null,[],[25,26,27],"regression","ml-models","tabular-data",[29],{"name":30,"url":31},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.22422",0]