On the Construction of Neuromolecular Chip: A Biologically Motivated Evolvable Hardware

摘要

Computer systems have fast processing speed for performing heavy computational tasks; however, it is lack of flexibility in dealing with most real world problems. Relatively, organisms have high adaptability in dealing with environmental changes or noise. Evolvable neuromolecular hardware, motivated from some biological evidence is proposed. The ability to evolve, self-organizing dynamics, and a closed structural-function relationship are the three principle features. The objective in this research is to capture these three biological features and implement them onto a digital circuit. The chip was designed with Verilog and applied to pattern recognition problem domain. Experimental result shows that the structures and functions of this hardware are closely related that its outputs change accordingly when we gradually modify the structure, implying this hardware embracing an adaptive fitness landscape that facilitates evolutionary learning.