The focus of this project is to investigate and develop methodologies and approximation schemes for the efficient mapping of probabilistic models such as Bayesian Neural Networks on low-power embedded devices such as CPUs, NPUs, and FPGAs. The project will be mainly based in the Department of Electrical and Electronic Engineering within the Circuits and Systems Group (http://www.imperial.ac.uk/electrical-engineering/research/circuits-and-systems/
), at South Kensington Campus, where the student will have the opportunity to spend some time of the PhD in Arm, Cambridge.
Despite the significance progress of neural network acceleration, it is well known that conventional neural networks can be prone to overfitting and poor generalisation – where the model fails to generalise well from the training data to test (unseen) data. The fundamental reason is that traditional deep neural network models do not provide estimates with uncertainty information which is otherwise very critical for real-life scenarios – e.g. autonomous driving, healthcare etc. Bayesian Neural network and Gaussian Process Models can easily learn from small datasets, with the ability to offer estimates of uncertainty and robustness to mitigate poor generalisation issues. The goal of this research is to improve our understanding of the computational aspects of these types of probabilistic models and to explore various optimisation strategies across the full stack (from model to hardware) for efficient implementation of them on low-power embedded platforms – e.g. CPU, NPU etc.
The PhD student will be jointly supervised by Dr Christos-Savvas Bouganis (Imperial College) and Dr Partha Maji (Arm).
A competitive candidate for this role should demonstrate the following:
• A good First-Class Degree (or International equivalent), in electronics or computer science
• A Masters level degree qualification and/or relevant experience.
Experience and skills:
• Software development for deep learning
• Computer arithmetic
• Programming embedded systems
• FPGA circuit design
• Probabilistic machine learning
A lack of experience in the above experience and skills could be compensated by evidence of research potential.
For queries regarding academic or technical skills please contact Dr Bouganis ([email protected]
Please click here to apply (https://imperialuk.elluciancrmrecruit.com/Apply/Account/Login?ReturnUrl=%2fApply%2f
). Course code: Electrical Engineering Research – H6ZX
NB: In the application, the proposed supervisor should be “Dr Christos-Savvas Bouganis”.
Any queries regarding the application should be directed to Lina Brazinskaite [email protected]
Closing Date: 30 June 2020 - However, applications will continue to be accepted until the position is filled.
About Imperial College London
Imperial College London is the UK’s only university focussed entirely on science, engineering, medicine and business.
Additional information on the PhD programme in the Department of Electrical and Electronic Engineering can be found here.