Support Vector Regression (fitrsvm) with prior knowledge constraint

1 次查看（过去 30 天）
Dave Lee
Dave Lee 2016-11-30
编辑： Dave Lee 2016-11-30
Hi there, I am using fitrsvm to estimate a regression model. The model fits data very well and overfitting is guaranteed since I use cross-validation. However, given the fact that the data is not rich enough to estimate the real physics model, I obtained the model which violates some physical constraints, i.e. partial derivative of the estimated function vs. independent variables must be positive/negative. I went through couple literature, such as https://hal.archives-ouvertes.fr/hal-00178619/document and see that this is a resolvable problem. Could anyone share with me how to incorporate these constraints into fitrsvm? Thanks a lot, Dave

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Support Vector Machine Regression 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by