next up previous
Next: Regularization Up: Measures Against Overfitting Previous: Measures Against Overfitting

Order Reduction

One way of dealing with overfitting is by reducing the order of the transformed data $ \textbf{\~X}_k$ through PCA or a similar technique [11,12,13], in which the data $ \textbf{\~X}_k \in \mathbb{R}^{N \times m_1'}$ are to be transformed to $ \textbf{\=X}_k \in \mathbb{R}^{N \times m_k''}$ with $ m_k'' < N$ by considering its $ m_k''$ principal components.



Steven Van Vaerenbergh
Last modified: 2006-04-05