Sokolov E N, Nezlina N I
M. V. Lomonosov Moscow State University, Moscow, Russia.
Neurosci Behav Physiol. 2010 Mar;40(3):279-93. doi: 10.1007/s11055-010-9255-y. Epub 2010 Feb 11.
Grouping, segmentation, and accentuation - processes involved in stimulus perception - are discussed. These effects are explained in terms of the universal vector coding model in neural networks. Grouping is the combination of objects or events into units on the basis of their similarity. Segmentation, conversely, is the separation of groups to the level of ensembles consisting of small numbers of objects. The processes of grouping and segmentation are regarded from the point of view of their underlying neural mechanisms. It is suggested that stimuli in neural networks are encoded by patterns of excitation of cardinal neurons. These excitation patterns can be represented as excitation vectors. Differences between stimuli are formed as the absolute magnitudes of their vector differences. The greater the perceived stimuli differ from each other, the greater the difference in their perceptual and semantic excitation vectors. The more similar the stimuli, the smaller their vector difference. This suggests that stimuli with similar excitation vectors will be grouped together in perceptual space. Conversely, stimuli with different excitation vectors will "repel" and become segmented. The spatial separation of objects increases with increases in the differences between their spatial excitation vectors. The universality of the vector coding principle can be illustrated using color contrast as an example: differences in contrasting colors increase with increases in the differences between their excitation vectors. Groups of objects with similar excitation vectors are accentuated in perception by means of summation of their excitation vectors. Groups of objects with different excitation vectors undergo mutual accentuation because of the appearance of contrast. Plastic accentuation is associated with the novelty of stimuli and is extinguished on repetition of the stimulus.
讨论了分组、分割和突显——刺激感知过程中涉及的过程。这些效应是根据神经网络中的通用矢量编码模型来解释的。分组是基于对象或事件的相似性将它们组合成单元。相反，分割是将组分离到由少量对象组成的集合级别。从其潜在神经机制的角度来考虑分组和分割过程。有人提出神经网络中的刺激是由主要神经元的兴奋模式编码的。这些兴奋模式可以表示为兴奋矢量。刺激之间的差异表现为它们矢量差异的绝对大小。感知到的刺激彼此差异越大，它们的感知和语义兴奋矢量的差异就越大。刺激越相似，它们的矢量差异就越小。这表明具有相似兴奋矢量的刺激将在感知空间中被分组在一起。相反，具有不同兴奋矢量的刺激将“相互排斥”并被分割。对象的空间分离随着它们空间兴奋矢量差异的增加而增加。矢量编码原理的普遍性可以用颜色对比作为例子来说明：对比色之间的差异随着它们兴奋矢量差异的增加而增加。具有相似兴奋矢量的对象组通过它们兴奋矢量的总和在感知中被突显。具有不同兴奋矢量的对象组由于对比的出现而相互突显。可塑性突显与刺激的新颖性相关，并在刺激重复时消失。
