Module kpca 

Kernel PCA for nonlinear dimensionality reduction Kernel PCA (Principal Component Analysis)

Kernel PCA is a nonlinear extension of PCA that uses the kernel trick to perform PCA in a high-dimensional (potentially infinite-dimensional) feature space without explicitly computing the feature map.

Algorithm

1. Compute the kernel (Gram) matrix K
2. Center K in feature space
3. Eigendecompose the centered K
4. Project data using the top eigenvectors

Features

• Nonlinear dimensionality reduction via kernel trick
• Pre-image estimation (approximate reconstruction to input space)
• Automatic kernel parameter selection via grid search on reconstruction error
• Support for all kernel types in the kernels module

Structs

KernelPCA

Kernel PCA for nonlinear dimensionality reduction