Learning multiple feature representations from natural image sequences

Abstract

Hierarchical neural networks require the parallel extraction of multiple features. This raises the question how a subpopulation of cells can become specific to one feature and invariant to another, while a different subpopulation becomes invariant to the first but specific to the second feature. Using a colour image sequence recorded by a camera mounted to a cat's head, we train a population of neurons to achieve optimally stable responses. We find that colour sensitive cells emerge. Adding the additional objective of decorrelating the neurons' outputs leads a subpopulation to develop achromatic receptive fields. The colour sensitive cells tend to be non-oriented, while the achromatic cells are orientation-tuned, in accordance with physiological findings. The proposed objective thus successfully separates cells which are specific for orientation and invariant to colour from orientation invariant colour cells. © Springer-Verlag Berlin Heidelberg 2002.