James Malkin

PhD student

with Laurence Aitchison and Conor Houghton

As a Bristol Neuroscience undergraduate I worked with Cian on a summer project where I gained insight into the importance of variation in intracellular calcium dynamics in encoding information, through which inference of plasticity expression is determined. My Neuroscience dissertation focussed on a relatively neglected topic, neurovascular coupling and energetics. These projects were to be the bedrock of my present PhD proposal. After Bristol, I completed a Computer Science MSc at UCL, in which I most enjoyed machine learning and neural network modules. With a rationale rooted in my interests in neuro-energetics and learning, the theme of my MSc summer project was to demonstrate efficiency gains of learning algorithms that enact increasingly sophisticated statistical inference on inputs. The MSc project concluded that learning through variational Bayes admitted vast gains in efficiency.

I am now working with Cian, Laurence Aitchison, and Conor Houghton to demonstrate with rigour that concerning the condition of metabolic energy minimisation, learning must approximate variational Bayesian inference.