Associative memory model that extracts automatically statistical information included in learning-order and recognizes objects in self organizing

点击翻译按钮获取中文摘要

If the observing order of observing patterns is always fixed, it might be the pattern in which same object is observed from various direction. If the randomness of the observing order is strong, the possibility that the patterns are independent mutually is high. This study examined the dependence of the attractor structure on the statistical properties of the learning order in the Amit model modified Hopfield model. We found that if the statistical properties of the learning order change, stable state can change to the appropriate attractor reflecting the relationship between memory patterns.