3

With PCA, it is possible to reconstruct high-dimensional data from its low-dimensional point by $$ x_i' = Pb + \bar X $$ Where $\bar X$ is the mean of training set $X$, $P$ is the eigenvectors and $b$ is a principal component.

Can this be achieved with kernel PCA (KPCA)? If so, how?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Bangau
  • 31
  • 2
  • 2
    +1. See my answer here http://stats.stackexchange.com/questions/131142, in particular around the reference to the Mika et al. 1998 paper. They call it "pre-image". You can also try googling `pre-image kernel PCA` (try this [google scholar query](https://scholar.google.com/scholar?hl=en&q=pre-image+kernel+pca)) and browse through the literature. I am not particularly familiar with it. – amoeba Jun 22 '16 at 13:09

0 Answers0