Geoffrey designs machine learning algorithms. His aim is to discover a learning procedure that is efficient at finding complex structures in large, high-dimensional datasets, and to show that this is how the brain learns to see.