University of Twente Student Theses

Login

Finding intelligence in disordered boron dopant atoms

Nass, M. (2018) Finding intelligence in disordered boron dopant atoms.

[img] PDF
Restricted to Restricted

4MB
Abstract:In this paper we try to make the first steps towards materializing neural networks. Materializing a neural network will very likely improve speed and power efficiency. Our group tries to lay a basis for neuromorphic computers and find a more complex behaviour than we have seen before by undesigned chips. Using a low boron dopant concentration in silicon we create a chip that has very interesting exploitable properties, for example highly none linear IV-curves (at 77K). In this thesis we investigate the applicability of this construct to solve a policy problem. We find that the chip has the ability to distinguish situations that look fairly similar and that using our chip as a kernel trick the policy problem can be solved. The chip also shows promising results regarding solving the described policy problem directly. By solving the policy problem in the kernel trick like way we give a proof of concept for future implementations like feature extraction and maybe even direct number recognition. In addition, we study the possibility of capturing the functionality of such a chip with neural networks. We find similar behaviour fitting the neural network to synthetic data compared to measurements. These results can guide us to improve the learning performance on measurement data.
Item Type:Essay (Bachelor)
Faculty:TNW: Science and Technology
Subject:31 mathematics, 33 physics, 50 technical science in general, 51 materials science, 53 electrotechnology, 54 computer science
Programme:Applied Physics BSc (56962)
Link to this item:http://purl.utwente.nl/essays/76389
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page