University of Twente Student Theses


Visual classification with a simulated disordered boron dopant network

Liu, Jo-Yu (2021) Visual classification with a simulated disordered boron dopant network.

[img] PDF
Abstract:As artificial intelligence slowly approaches human-like intelligence and capabilities, limitations to current computational technology and architecture surfaces, and creates barriers and challenges towards the scaling of neural systems that are inherently too large and complex to simulate. With millions of neurons in V1 alone, a prominent, and concurrent example of the affected domains is machine vision. By harnessing the inherent, in-material, complexity of a boron dopant atoms network, Chen et al’s (2020) neuromorphic hardware research has brought about a promising solution to the ever increasing computational needs that machine vision artificial intelligence demands. The present paper sets out to simulate, and find out how feasible it is to use, the boron network to solve Boolean logic gates and to run current machine vision algorithms that mimic human visual perceptual architectures. The boron dopant cells are simulated as reservoir networks, using perceptron learning and later, a self-interpreted control node scheme. The results show that the boron-reservoir architecture introduces a form of complexity between the input and output, which, with its inherent randomness, reduces the reliability of all simulated networks in its ability to solve simple problems like the Boolean logic gates, but also provides enough complexity to sometimes solve the nonlinear logic gates. By altering the cell and network architecture, the inherent randomness can be reduced so that functional cells can be reliably created. The simulated architecture is then extended to simulate basis pattern recognition given by digit recognition and by simulating so-called simple cells as found in the first two layers of a model based on the visual cortex.
Item Type:Essay (Master)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:77 psychology
Programme:Psychology MSc (66604)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page