A method for room temperature demonstration of in-materio reservoir computing (RC) with a single-walled carbon nanotube/porphyrin-polyoxometalate network (SWNT/Por-POM) is proposed. Boolean functions of OR, AND, NOR, NAND, XOR, and XNOR, all were reconstructed with an accuracy >90% via supervised training of linear voltage readouts. The RC pre-requisite of echo-state property and recurrent connection allowed for consistent performances over multiple test datasets and time-shifted target sequences. Moreover, a non-zero machine intelligence index confirmed the presence of negative differential resistance dynamics, incorporating in SWNT/Por-POM the mathematical equivalence of additive and subtractive functions, thereby aiding the construction of such complex Boolean functions.
ASJC Scopus subject areas
- General Engineering
- General Physics and Astronomy