University of Twente Student Theses

Login

Dopant networks for energy-efficient classification

Wilde, B. de (2019) Dopant networks for energy-efficient classification.

Full text not available from this repository.

Full Text Status:Access to this publication is restricted
Abstract:Recently, it has been shown that a network of boron dopants in silicon can solve various classification problems in materio, including the canonical linearly inseparable XOR and XNOR. Here, a kinetic Monte Carlo modelling approach, addressing charge transport in the variable-range hopping (VRH) regime, is shown to replicate such networks numerically. The model is verified against experiment by comparing current-voltage characteristics and the temperature dependence of the resistance obtained for actual devices. Simulated devices are shown to be able to solve each of the basic Boolean logic gates, including XOR and XNOR. The voltage configurations for each logic gate are found by means of a genetic algorithm. One central promise of the field of brain-inspired hardware is that it will lower the energy consumption of computation. This report presents approaches towards realising this goal, both numerically and experimentally. Experimentally, energy-efficient computation was realised by incorporating power consumption in the fitness function of the genetic algorithm used to find logic gates, leading to two times lower static power consumption on average. Numerically, a scheme is proposed that uses electrostatically defined electrodes, leading to orders of magnitude lower static power consumption compared to the original approach in three out of six gates.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:31 mathematics, 33 physics
Programme:Applied Mathematics MSc (60348)
Link to this item:https://purl.utwente.nl/essays/78009
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page