Contact
Biography
I'm a computational neuroscientist and applied mathematician studying how energy constraints shape learning in biological and artificial systems. I completed my PhD at the University of Bristol, where I developed models linking synaptic variability, resource allocation and learning. My broader interests include statistical mechanics, Bayesian inference and the mathematical principles of neural systems. At Nottingham, my postdoctoral work develops a framework that combines the functional benefits of learning with its state-dependent energetic costs, aiming to identify learning rules that make the most progress for a given energy budget. I will test these ideas in neural models and experimental data, and explore their application to energy-efficient AI and neuromorphic systems.
Expertise Summary
My expertise spans computational neuroscience, probabilistic machine learning and mathematical modelling. I develop stochastic models of synaptic transmission and plasticity, using constrained optimisation and variational Bayesian inference to connect metabolic resource allocation with synaptic precision and task performance. My work includes quantal models of release probability and amplitude, energy-precision tradeoffs, and discrete, mean-field and continuous descriptions of learning dynamics. .
My current work develops energy-aware learning rules that combine Fisher information and natural gradients with state-dependent energetic costs, represented through convex potentials and Bregman distances. I use mathematical analysis, neural-network simulation and experimental data to investigate these ideas, with applications to biological learning, energy-efficient AI and neuromorphic systems.
Research Summary
My current research examines how energy constraints shape learning in biological and artificial systems. At Nottingham, I am developing geometric approaches that combine the functional sensitivity of… read more
Current Research
My current research examines how energy constraints shape learning in biological and artificial systems. At Nottingham, I am developing geometric approaches that combine the functional sensitivity of a learning system with the energetic costs of changing its parameters. Alongside this, I model synaptic plasticity as resource investment followed by relaxation, predicting changes in transmission variability and the contributions of release probability and quantal amplitude. I also investigate how biological mechanisms that suppress fluctuations relate to adaptive optimisation algorithms.
Past Research
My doctoral research at the University of Bristol investigated the energetic costs of reliable synaptic transmission. I incorporated physiological cost models into neural networks with stochastic synapses, showing how resource allocation shapes synaptic variability and task performance. This work established a formal connection with variational Bayesian inference. I subsequently studied how release probability, release-site number and quantal amplitude implement synaptic precision, deriving resource-precision tradeoffs and testing their predictions against experimental measurements.
Future Research
I aim to develop a general framework for efficient learning when energetic costs vary with system state and are coupled through shared resources. This will connect information geometry with energetic constraints and produce learning rules that use locally available activity and metabolic signals. I plan to test the resulting predictions in mechanistic neural models and experimental studies of synaptic plasticity and metabolism, then explore their application to artificial intelligence and neuromorphic computing.