{"id":"arxiv/2605.01446","url":"https://arxiv.org/abs/2605.01446","tags":["math.NA","stat.ML"],"score":0,"title":"Sequential Minimal Optimization for $\\varepsilon$-SVR with MAPE Loss and Sample-Dependent Box Constraints","domain":"math.NA","source":"arxiv","authors":["Pablo Benavides-Herrera","Riemann Ruiz-Cruz","Juan Diego Sánchez-Torres"],"reviews":0,"subject":"math.NA","summary":"Support vector regression with Mean Absolute Percentage Error (MAPE) loss is theoretically well-motivated for forecasting applications where accuracy is evaluated in relative terms, but the sample-dependent dual box constraints it induces have not been addressed in the published SMO literature. We derive a Sequential Minimal Optimization algorithm for this setting and prove a structural-invariance result: the MAPE modification affects exactly two components of the SMO iteration -- working-set selection and analytic-update clipping -- leaving gradient bookkeeping and curvature computation…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-05-02","reactions":{}}