Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Superconducting neural network using quantum-flux parametrons
Thesis (MEng)--Stellenbosch University, 2023.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | en_ZA en_ZA |
| Published: |
Stellenbosch : Stellenbosch University
2023
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867614019042410496 |
|---|---|
| access_status_str | Open Access |
| author | Jardine, Muhammad Ameen |
| author2 | Fourie, Coenrad |
| author_browse | Fourie, Coenrad Jardine, Muhammad Ameen |
| author_facet | Fourie, Coenrad Jardine, Muhammad Ameen |
| author_sort | Jardine, Muhammad Ameen |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (MEng)--Stellenbosch University, 2023. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/127205 |
| institution | Stellenbosch University (South Africa) |
| language | en_ZA en_ZA |
| last_indexed | 2026-06-10T12:45:22.846Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2023 |
| publishDateRange | 2023 |
| publishDateSort | 2023 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/127205 Superconducting neural network using quantum-flux parametrons Jardine, Muhammad Ameen Fourie, Coenrad Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering. Superconductivity Neural computers Algorithms Thesis (MEng)--Stellenbosch University, 2023. ENGLISH ABSTRACT: This thesis presents the design of a simple superconducting neural network using Quantum-Flux Parametrons (QFPs) and Rapid Single-Flux Quantum Logic (RSFQ) circuits. Neural networks are systems that mimic the brain’s algorithms and learning techniques in order to recognise and respond to stimuli. Neural networks are ubiquitous and are used in a number of sectors, from finance to engineering, and are implemented mostly in software. However, as the improvement of traditional CMOS technology begins to slow, improvements in the operation of neural networks have begun to bottleneck. Neuromorphic computing aims to provide a solution to this bottleneck, by designing analogue circuits that mimic the brain’s algorithms, as opposed to having software implement them on generic hardware. Superconducting electronics provide an interesting approach to neuromorphic computing through superior speeds and efficiency. Th e pr oposed ne uromorphic sy stem co nsists of tu nable synaptic circuits, activation circuitry, multi-fluxon storage and implementations of traditional neural network algorithms, using Boolean logic circuits. The process used to design the circuitry is tested using SPICE simulation for various parameters in order to find t he most f easible d esign i n t erms o f b ehaviour and practicality. Algorithms are tested in scripting software before being emulated using the superconducting logic gates. Simulations suggest that the designed neural network system is capable of learning at relatively high speeds and from varying random starting points. This is promising for future iterations of the system, whereby more complex learning tasks could be achieved. AFRIKAANS OPSOMMING: Hierdie tesis bied die ontwerp aan van ’n eenvoudige supergeleidende neurale netwerk wat gebruik maak van Quantum-Flux Parametrons (QFP’s) en Rapid Single-Flux Quantum Logic (RSFQ) stroombane. Neurale netwerke is stelsels wat die brein se algoritmes en leertegnieke naboots om stimuli te herken en daarop te reageer. Die gebruik van neurale netwerke is wydverspreid in verskeie sektore, van die finansiële sektor tot ingenieurswese, en word meestal in sagteware geïmplementeer. Die tempo van verbetering van tradisionele CMOS-tegnologie begin egter afneem, wat verbeterings in die werking van neurale netwerke hinder. Neuromorfiese r ekenaars kan g ebruik word om ’n oplossing vir hierdie hindernis te bied by wyse analoogstroombane te ontwerp wat die brein se algoritmes naboots, eerder as sagteware-implementasies op generiese hardeware. Die hoër spoed en effektiwiteit van supergeleidende elektronika gee aanleiding tot ’n interessante benadering vir neuromorfiese berekening. Die voorgestelde neuromorfiese s telsel b estaan uit i nstelbare sinaptiese stroombane, aktiveringskringe, multi-fluxonberging, e n implementering van tradisionele neurale-netwerkalgoritmes, deur Boolese-logikastroombane te gebruik. Die stroombaan-ontwerpsproses word getoets met behulp van SPICEsimulasies oor verskeie parameters, ten einde die mees lewensvatbare en praktiese ontwerp in terme van stroombaan gedrag te verkry. Algoritmes is in skripsagteware getoets voordat hul gedrag in supergeleidende logika-hekke gewaarword is. Simulasies dui daarop dat die ontwerpte neurale-netwerkstelsel die vermoë het om teen relatiewe hoë snelhede en vanaf verskillende lukrake beginpunte te kan leer. Hierdie sien belowend uit vir toekomstige iterasies van die stelsel, waarin meer komplekse leertake bereik kan word. Masters 2023-03-03T11:37:13Z 2023-05-18T07:09:43Z 2023-03-03T11:37:13Z 2023-05-18T07:09:43Z 2023-03 Thesis http://hdl.handle.net/10019.1/127205 en_ZA en_ZA Stellenbosch University xiv, 164 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Superconductivity Neural computers Algorithms Jardine, Muhammad Ameen Superconducting neural network using quantum-flux parametrons |
| title | Superconducting neural network using quantum-flux parametrons |
| title_full | Superconducting neural network using quantum-flux parametrons |
| title_fullStr | Superconducting neural network using quantum-flux parametrons |
| title_full_unstemmed | Superconducting neural network using quantum-flux parametrons |
| title_short | Superconducting neural network using quantum-flux parametrons |
| title_sort | superconducting neural network using quantum flux parametrons |
| topic | Superconductivity Neural computers Algorithms |
| url | http://hdl.handle.net/10019.1/127205 |
| work_keys_str_mv | AT jardinemuhammadameen superconductingneuralnetworkusingquantumfluxparametrons |