University of Twente Student Theses
An Oscillatory Neural-Network architecture based on Dopant network Processing Units
Buitenweg, M. (2023) An Oscillatory Neural-Network architecture based on Dopant network Processing Units.
|
PDF
836kB |
| Abstract: | The amount of data is growing faster than ever, yet the crucial task of comprehending and deriving meaning from this information persists. This seemingly insatiable demand for inference is to be met within an obviously bounded energy budget, that must be made to stretch as far as possible. Consequently, alternative computing paradigms are being investigated with the aspiration of surpassing current energy-efficiency standards and achieving higher performance. Among these paradigms are the Dopant Network Processing Units (DNPUs), a class of high-dimensionally tunable non-linear silicon-based devices that have recently been shown to be capable of energy- and footprint-efficient compute. Oscillatory Neural Networks (ONNs) form a less recent computing paradigm aiming at raising efficiency by exploiting the synchronization phenomena found in oscillator networks to compute in phase and frequency instead of amplitude. In this BSc Assignment, we investigate ways of utilizing DNPUs to create an enhanced ONN architecture. A phase-computing DNPU-based ONN architecture is described. Then, an ONN simulator is developed and used to demonstrate that the described architecture can perform associative memory and classification tasks. It is shown that these tasks can also be performed when the coupling between oscillators is nonlinear, especially if the network complexity is low. The set of tasks shown in this work leave the computational capabilities of the DNPUs in the ONN underutilized, leaving room for further investigation. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 53 electrotechnology |
| Programme: | Electrical Engineering BSc (56953) |
| Link to this item: | https://purl.utwente.nl/essays/96272 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page