In preceding chapters, we considered only linear features. In this chapter, we introduce non-linear features. There is an infinite variety of different kinds of non-linearities that one might use for computing such features. We will consider a very basic form inspired by models previously used in computer vision and neuroscience. This approach is based on the concepts of subspaces and energy detectors. The resulting model called “independent subspace analysis” gives non-linear features which turn out to be very similar to complex cells when the parameters are learned from natural images.
CITATION STYLE
Hyvärinen, A., Hurri, J., & Hoyer, P. O. (2009). Energy Detectors and Complex Cells (pp. 213–237). https://doi.org/10.1007/978-1-84882-491-1_10
Mendeley helps you to discover research relevant for your work.