Assembly pointers for variable binding in networks of spiking

5 Currently existing neural network models lack an important computational capability of 6 networks of neurons in the brain: the capability to reason on an abstract level, and to structure 7 information according to abstract categories. While there is hardly anything known about 8 underlying processes for that in the human brain, experimental evidence for the organization 9 of an important prerequisite for abstract reasoning has recently been found: for binding words 10 to syntactic categories, such as subject or object. Based on these experimental data from fMRI 11 recordings we propose a model for binding of semantic content to variables that is consistent 12 with these data and provides new functions to neural network computations. We provide a 13 proof of principle that this model can be implemented even in networks of spiking neurons. 14 It enables structured information retrieval and copying of information without routing, and 15 provides a basis for the emulation of more demanding cognitive computations in networks of 16 spiking neurons. [February 2, 2018, 12:04] 17