Neuro-inspired learning of low-level image processing tasks for implementation based on nano-devices
As the fabrication cost of CMOS mask increases exponentially while the technology is approaching its physical limits, research interest focuses on emerging technologies and alternative architectures. Non-volatile components are considered as possible alternative technologies and neural networks constitute an interesting framework. Here, we present a learning strategy applied to a new non volatile device: the Optically Gated Carbon Nanotube Field Effect Transistor (OG-CNTFET). In this paper, electrical simulations using accurate compact model demonstrate the efficiency of this method to learn linearly separable Boolean functions.