Design and Simulation of Spiking Programmable Neural Computers
Authors: Mehmet Kerem Turkcan
Presentation type: Poster
We consider the design of spiking neural machines capable of implementing Turing machines, and using the design principles considered we propose a programmable spiking neural computer architecture. In the computer, specific subcircuits openly represent instructions. We extend the proposed architecture to allow for task parallelism via the fork-join model. In addition to the design, we implement a Python-based simulator for our computer that can be executed efficiently on GPUs. The approach proposed explicitly gives a computational meaning to each neuron in terms of the instruction it implements or in regards to the variable register it represents.