THE INDEPENDENT COMPONENTS OF NATURAL SCENES ARE EDGE FILTERS

It has previously been suggested that neurons with line and edge selec tivities found in primary visual cortex of cats and monkeys form a spa rse, distributed representation of natural scenes, and it has been rea soned that such responses should emerge from an unsupervised learning algorithm that attempts to find a factorial code of independent visual features, We show here that a new unsupervised learning algorithm bas ed on information maximization, a nonlinear ''infomax'' network, when applied to an ensemble of natural scenes produces sets of visual filte rs that are localized and oriented, Some of these filters are Gabor-li ke and resemble those produced by the sparseness-maximization network. In addition, the outputs of these filters are as independent as possi ble, since this infomax network performs Independent Components Analys is or ICA, for sparse (super-gaussian) component distributions, We com pare the resulting ICA filters and their associated basis functions, w ith other decorrelating filters produced by Principal Components Analy sis (PCA) and zero-phase whitening filters (ZCA), The ICA filters have more sparsely distributed (kurtotic) outputs on natural scenes, They also resemble the receptive fields of simple cells in visual cortex, w hich suggests that these neurons form a natural, information-theoretic coordinate system for natural images. (C) 1997 Elsevier Science Ltd.