Deutsch S
Biol Cybern. 1981;42(1):17-21. doi: 10.1007/BF00335154.
In a recent paper, Kunihiko Fukushima described a "Neocognitron: a self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position". The present paper presents a simplified version of the neocognitron. Whereas the latter employs a 16 x 16-element visual field which requires computer simulation, the simplified model uses a 10-element one-dimensional visual field. Two input examples are analyzed: a white sheet which is gradually lowered over a black background, and a white center dot which gradually stretches vertically in both directions until it covers the black background. The model demonstrates invariance with respect to lateral shift.
在最近的一篇论文中，福岛邦彦描述了一种“新认知机：一种用于模式识别机制的自组织神经网络模型，不受位置偏移的影响”。本文提出了新认知机的一个简化版本。后者采用一个16×16元素的视野，这需要计算机模拟，而简化模型使用一个10元素的一维视野。分析了两个输入示例：一张在黑色背景上逐渐下降的白纸，以及一个白色中心点，它在两个方向上垂直逐渐伸展，直到覆盖黑色背景。该模型展示了相对于横向偏移的不变性。
