Bio
Charles Jones is a PhD student at Imperial College London, working on machine learning for medical imaging. His research draws on the mathematical field of causality to understand and overcome challenges when translating methods into clinical practice. He works to develop machine learning methods which will be safe, robust, and fair in deployment.
Before starting his PhD in Computer Science, Charles was awarded an MEng in Mechanical Engineering, where he worked on prosthetic devices and neurosurgical robotics. Today, he remains interested in cross-disciplinary opportunities involving developing and applying technology for health.
Research interests
Deep learning methods exhibit impressive performance on image analysis tasks, such as classifying X-rays with radiologist-level performance. However, today's methods can reflect or amplify bias in their training data and often suffer performance degradation when deployed in the real world. These issues damage the credibility of deep learning methods, deterring adoption in high-stakes settings such as medical diagnosis.
In recent years, promising avenues for tackling these issues have drawn on the field of causal reasoning, which provides a principled mathematical formulation of the problems. Charles' research at the Alan Turing Institute contributes to the emerging field of 'causal deep learning', which combines the rigour and theoretical insights of causality with the effectiveness of modern deep learning. He aims to develop deep learning methods that are transparent and robust to spurious correlations. Instead of reflecting and amplifying biases in training data, Charles will develop causal methods to identify and mitigate them.