The Fixed-Point Algorithm and Maximum Likelihood Estimation forIndependent Component Analysis
Neural Processing Letters
A Learning Rule to Model the Development of Orientation Selectivity in Visual Cortex
Neural Processing Letters
Hi-index | 0.00 |
Scientists and experts have explored the mechanism of visual systems for decades for smart image processing and pattern recognition in order to satisfy sophisticated engineering applications. In this paper we apply independent component analyses' (ICA) unsupervised learning to natural images, topography images and other special environment images for demonstrating the simple-cell's process in animal vision. Our results confirm an early biological experiment (Nature 228 (1970) 419) about the growth of simple cells in cat's V1 area. Furthermore, by applying ICA methodology and the simplex algorithm, the unsupervised neural synapse's learning can obtain the receptive fields in visual cortex and can simulate the growth of the visual cortex of young animal in the special environment. These findings imply that an input image can be efficiently represented by ICA bases. An application of image matching in the navigation by ICA is shown that the animal visual system method is indeed better than those classical methods at least more than 5%.