Kernel trick

The kernel trick evaluates feature-space inner products through $k(x,y)$ without explicitly constructing the feature vectors.

Kernel principal component analysis diagonalizes a centered kernel matrix to perform principal component analysis in an implicit feature space.