]> BibSonomy publications for/author/Field Nature607Emergence of Simple-Cell Receptive Field Properties By Learning A Sparse Code for Natural Images.3811996vis 2008-09-16 23:39:07.0 Current Opinion in Neurobiology4481-487Sparse coding of sensory inputs142004vis 2008-09-16 23:39:07.0 Journal of Experimental Psychology: Human Perception and Performance815Perceptual-Motor Sequence Learning of General Regularities and Specific Sequences.251999cogn 2008-09-16 23:39:07.0 Science179-181Discrimination and imitation of facial expressions by neonates2181982devo, vis 2008-09-16 23:39:07.0 Infant behavior and development19-25Mother-stranger face discrimination by the newborn71984devo, vis 2008-09-16 23:39:07.0 Neural Computation4559-601What Is the Goal of Sensory Coding?61994nnets 2008-09-16 23:39:07.0 Learning and Motivation446-464Conceptual Conditioning: Evidence for an Artifactual Account of Evaluative Learning281997imported 2008-09-16 23:39:07.0 KDDconf/kdd/2008301-309Quantitative evaluation of approximate frequent pattern mining algorithms.2008dblp 2008-09-02 00:00:00.0 New York, NY, USA192--201Semantics-based reverse engineering of object-oriented data models2006imported 2008-07-07 16:45:32.0 Nature6583607--609{Emergence of simple-cell receptive field properties by learning a sparse code for natural images}3811996featureextraction receptivefields 2008-06-12 08:20:57.0The receptive fields of simple cells in mammalian primary visual cortex can be characterized as being spatially localized, oriented and bandpass (selective to structure at different spatial scales), comparable to the basis functions of wavelet transforms. One approach to understanding such response properties of visual neurons has been to consider their relationship to the statistical structure of natural images in terms of efficient coding. Along these lines, a number of studies have attempted to train unsupervised learning algorithms on natural images in the hope of developing receptive fields with similar properties, but none has succeeded in producing a full set that spans the image space and contains all three of the above properties. Here we investigate the proposal that a coding strategy that maximizes sparseness is sufficient to account for these properties. We show that a learning algorithm that attempts to find sparse linear codes for natural scenes will develop a complete family of localized, oriented, bandpass receptive fields, similar to those found in the primary visual cortex. The resulting sparse image code provides a more efficient representation for later stages of processing because it possesses a higher degree of statistical independence among its outputs.