Overview: I am looking for a PhD student with an interest in the design of efficient and robust machine learning and data processing algorithms using tools from applied probability, signal processing and statistical mechanics. We will collaborate with applied researchers to apply the designed algorithms in the field of brain sciences.
Learning, and more generally information processing and storing, seems to work extremely well in biological systems. A bee (an animal with less than with less than a million neurons, and fewer than 10 billion synapses) can perform remarkably sophisticated optimisation of its behavior for the future. Machine learning algorithms have also achieved remarkable performance in recent years, however they are not flawless. For example, research has demonstrated that it can be easy to fool the algorithm into misreading a stop sign simply by putting four small stickers on the face of the sign. It is further estimated that the training of a model like GPT-3 may cost millions of dollars. This raises the question of whether we can define more efficient and robust learning models and algorithms. In this project, we will focus on the design of algorithms that perform robustly in varying scenarios and look into the design of more efficient machine learning models and optimisation algorithms using a variety of tools from different tools such as applied probability, statistical mechanics, information theory, signal processing and partial differential equations.
My group collaborates with applied domains such as robotics, privacy & security, and brain sciences working on projects in optimal control, deep brain stimulation and privacy-preserving machine learning algorithms. In this project we will apply some of the designed algorithms and obtained insights into the above-mentioned areas (depending on your interests).
The position: You will be part of the Department of Mathematics as well as Imperial College London's new Imperial-X initiative. Imperial-X is a center focused on multidisciplinary theoretical and applied research in machine learning performed in close collaboration with industrial partners. The PhD project is fully funded with an additional budget for conferences and research visits. For more information on I-X please visit: https://www.imperial.ac.uk/stories/ix-rapid-innovation/
What am I looking for? I am looking for someone with strong technical skills in for example applied probability, signal processing, statistical mechanics or partial differential equations, with a degree in applied mathematics, theoretical physics or electrical engineering. Due to the multidisciplinary nature of this project it'd be great if you are interested in both the theory of machine learning as well as its applications. A big focus on Imperial-X is on industrial collaborations and the commercialisation of research - so working with companies or spinning the research out into a startup can all be explored.