Analysis of the Collective Properties of "Neural Networks" and Their Relation to the Structure and Concepts of Memory

模型神经网络的集体兴奋的数学结构将被检验，以寻找它们可能与“高级神经功能”的基本方面的相关性。对这一课题的研究需要对神经细胞之间的连接模式有更多的了解。由于这不能直接从神经生物学实验中提供，我们将使用联想记忆的情况，其中连接的性质被认为是由于外部世界提供的信息中的结构。集体属性的性质预计取决于基本神经回路的拓扑结构。根本的问题是，任何集体激发之间的相互作用是否具有有用的计算性质。

The mathematical structure of the collective excitations of model neural networks will be examined for the relevance which they may have to the essential aspects of "higher nervous function." An examination of this subject requires that much be known about the pattern of connections between nerve cells. Since this cannot be provided directly from neurobiological experiments, we will work with the case of associative memory, where the nature of the connections is believed to be due to the structure in the information provided by the external world. The nature of the collective properties is expected to depend on the topology of the underlying neural circuits. The fundamental question is whether the interactions between any collective excitations have useful computational properties.