Project Title: Modern Hardware Acceleration in Deep Learning - a Novel Approach and Analysis  
Student: Daudi Bandres 
Course: MSci in Computer Science with Artificial Intelligence  

Abstract: 
This dissertation investigates emerging hardware solutions for accelerating neural network 
computations, including developing a novel implementation of a machine learning accelerator 
that can implement a neural networks directly to achieve higher performance and energy efficiency 
compared to traditional CPU and GPU-based approaches. The project aims to provide a comprehensive 
review of the state-of-the-art in hardware encoding of neural networks, investigate the performance, 
power consumption, and energy efficiency of different hardware acceleration techniques and 
implementations, evaluate the trade-offs between hardware acceleration and software-based 
implementations, and propose novel approaches and techniques for hardware encoding of neural 
networks. The research involves conducting extensive experiments using appropriate benchmarks and 
datasets and collecting and analyzing data to evaluate the effectiveness of different approaches. 
The end goal is to provide a better understanding of the benefits and drawbacks of hardware 
acceleration for machine learning and to invesitgate a new method of encoding artificial intelligence 
models into hardware that can potentially improve their performance and energy efficiency.