PulseAugur
LIVE 08:08:54
research · [2 sources] ·
0
research

New SMO algorithm developed for epsilon-SVR with MAPE loss

Researchers have developed a new Sequential Minimal Optimization (SMO) algorithm for $\varepsilon$-SVR that incorporates Mean Absolute Percentage Error (MAPE) directly into the loss function. This novel approach modifies the standard $\varepsilon$-SVR by introducing sample-dependent box constraints, which affects feasibility sets and clipping bounds. An implementation of this algorithm is available in the open-source "psvr" R package. AI

Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →

IMPACT Introduces a novel optimization algorithm for regression models, potentially improving accuracy in specific forecasting scenarios.

RANK_REASON This is a research paper detailing a new algorithm for a statistical modeling technique.

Read on arXiv stat.ML →

COVERAGE [2]

  1. arXiv stat.ML TIER_1 · Pablo Benavides-Herrera, Riemann Ruiz-Cruz, Juan Diego S\'anchez-Torres ·

    Sequential Minimal Optimization for $\varepsilon$-SVR with MAPE Loss and Sample-Dependent Box Constraints

    arXiv:2605.01446v1 Announce Type: cross Abstract: We derive a Sequential Minimal Optimization (SMO) algorithm for the quadratic dual problem arising from $\varepsilon$-SVR~\cite{Vapnik1995, Drucker1997, Smola2004} modified to minimize the Mean Absolute Percentage Error (MAPE)~\ci…

  2. arXiv stat.ML TIER_1 · Juan Diego Sánchez-Torres ·

    Sequential Minimal Optimization for $\varepsilon$-SVR with MAPE Loss and Sample-Dependent Box Constraints

    We derive a Sequential Minimal Optimization (SMO) algorithm for the quadratic dual problem arising from $\varepsilon$-SVR~\cite{Vapnik1995, Drucker1997, Smola2004} modified to minimize the Mean Absolute Percentage Error (MAPE)~\cite{Makridakis1993, Hyndman2006} directly in the lo…