Hardware Acceleration for Deep Learning

Artificial Intelligence (AI) is becoming the default choice for the many of the applications in the industries such as image processing and pattern recognition. As a result, there has been a great deal of work on developing and implementing AI algorithms to achieve more accuracy in both industry and academia. However, the insufficient computing power soon becomes a huge challenge for those well-developed algorithms because of exponentially increasing demand of the computing ability. This project will provide an opportunity to tackle this challenge by using different level of optimizations from the computing theory level to the algorithm level. By the sharp computing theory, architecture and algorithm we developed in this project, we will enable AI applications on more power sensitive and computing ability limited devices.

Detailed objectives are as follows:

1.      Identify the bottleneck for AI and its implementation on different platform

2.      Develop appropriate computing theory, architecture, number system and quantization approach to improve the performance of AI applications.

3.      Optimizing architecture specifically for the most computing intensive part

4.      Developing novel net structures for applications

5.      Evaluating the new approaches

6.      Publish journal and conference papers and present the work both internally and externally.

Entry requirements:

Candidates should have a minimum of an upper second honours degree (2.1) (or equivalent) in Computer Science/Electronics Engineering/Mathematics. A master’s degree is desirable but not necessary. Candidates in other degrees related to Engineering or related fields would be considered.