A missing data approach to validate nonlinear PCA

Abstract

Many biological processes behave in a nonlinear way. Observations over time usually show a curved trajectory in the data space. To understand the dynamics of biological processes we have to identify and analyze the time trajectory. This can be done by using a nonlinear extension of principal component analysis (PCA) which provides a noise-reduced description of the curved data structure. To avoid over-fitting a careful control of the model complexity is required for which we need a strategy to validate unsupervised nonlinear methods [1].