Emergence of Phase- and Shift-Invariant Features by Decomposition of Natural Images into Independent Feature Subspaces
- 1 July 2000
- journal article
- Published by MIT Press in Neural Computation
- Vol. 12 (7), 1705-1720
- https://doi.org/10.1162/089976600300015312
Abstract
Olshausen and Field (1996) applied the principle of independence maximization by sparse coding to extract features from natural images. This leads to the emergence of oriented linear filters that have simultaneous localization in space and in frequency, thus resembling Gabor functions and simple cell receptive fields. In this article, we show that the same principle of independence maximization can explain the emergence of phase- and shift-invariant features, similar to those found in complex cells. This new kind of emergence is obtained by maximizing the independence between norms of projections on linear subspaces (instead of the independence of simple linear filter outputs). The norms of the projections on such “independent feature subspaces” then indicate the values of invariant features.Keywords
This publication has 15 references indexed in Scilit:
- Independent component analysis by general nonlinear Hebbian-like learning rulesSignal Processing, 1998
- A Fast Fixed-Point Algorithm for Independent Component AnalysisNeural Computation, 1997
- A class of neural networks for independent component analysisIEEE Transactions on Neural Networks, 1997
- Equivariant adaptive source separationIEEE Transactions on Signal Processing, 1996
- Emergence of invariant-feature detectors in the adaptive-subspace self-organizing mapBiological Cybernetics, 1996
- WHAT IS THE GOAL OF SENSORY CODINGNeural Computation, 1994
- Independent component analysis, A new concept?Signal Processing, 1994
- Normalization of cell responses in cat striate cortexVisual Neuroscience, 1992
- Learning Invariance from Transformation SequencesNeural Computation, 1991
- Unsupervised LearningNeural Computation, 1989