Radial projections for nonlinear feature extraction

In this work, two new techniques for nonlinear feature extraction are presented. In these techniques, new features are obtained as radial projections of the original measurements. Radial projections are a particular kind of second order transformations that show interesting properties: they capture the local structure of the data and reduce dramatically the number of parameters to estimate from O(d²) to O(d). This reduction allows the efficient use of combinatorial optimization techniques (hill-climbing, genetic algorithms, simulated annealing, etc.) to search for transformations in high-dimensional spaces.