Keeken, Luuk, van (2021) Improving the Learning Performance of a Simulated Memristor Neural Network Using Optimisation Algorithms. Bachelor's Thesis, Artificial Intelligence.
|
Text
bAI_2021_LuukvanKeeken.pdf Download (819kB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (122kB) |
Abstract
A memristor is a relatively novel electronic component which has the property that its resistance can be adjusted, and that it can retain that new resistance. In contrast to the von Neumann architecture used in most modern computers, this opens up the possibility of co-locating memory and computation, as such mimicking the way in which the brain works. This allows for a decrease of computation time, an increase of energy efficiency, and a more brain-like computational substrate for artificial neural networks. Previous research has explored the use of Nb-doped SrTiO$_3$ memristors in differential pairs as weights of a simulated spiking neural network. The simulated model was capable of showing adequate performance in learning transformations of periodically time-varying input signals. The current research builds on that work by exploring the use of optimisation techniques based on Simulated Annealing and Stochastic Gradient Descent with Momentum. The implemented optimisation methods were found to significantly increase the learning performance of the simulated model.
| Item Type: | Thesis (Bachelor's Thesis) |
|---|---|
| Supervisor name: | Tiotto, T.F. and Borst, J.P. |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Bachelor's Thesis |
| Language: | English |
| Date Deposited: | 07 Jul 2021 10:32 |
| Last Modified: | 07 Jul 2021 10:32 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/24934 |
Actions (login required)
 |
View Item |