This project proposes a novel mathematical framework of Bayesian computing utilising the advantage of novel information units, neural spikes and synapses, of the brain. It aims to stimulate multidisciplinary innovations in establishing a neuro-inspired theoretical framework of statistical optimisation, achieving a transformative and general-purpose methodology for efficient, flexible and robust computing of multi-task optimisation.
One of the driving forces in applied mathematics is the provision of a theoretical framework for understanding relationships between data generated by studies in diverse complex fields, which in turn stimulates developing new theories and methods in mathematics. In recent years, Brain Science, as one of the most demanding and highly interdisciplinary fields, has produced a large amount of data spanning different levels, from fundamental biological cells to complex human behaviours. It is noteworthy that neurons - the building blocks of the brain’s information-computing units - encode information through individual events, or ‘spikes’, the timescale of which can be recorded in milliseconds. However, variables of cognition and behaviour are continuous. The discrete nature of spikes, generated from the dynamical equation of continuous neuronal voltage change, presents a challenge to classical statistical theories, where the mixture of discrete and continuous variables is not well addressed1. Thus the key question is how to develop a novel theoretical framework that can dissect and understand the complex interactions between discrete neural spikes and continuous behaviour variables. it is timely and meaningful to unify the inference of probabilities and learning of parameters into a single framework. To fill the gap of knowledge, we will develop a novel theoretical framework to take into account the sampling of neural spikes and synapses, achieving a general-purpose theoretical framework for flexible multi-task optimisation.
The supervision team has expertise in mathematics, probability and statistics, machine learning, computational neuroscience, with regularly published research papers at the very top venues in related fields.
Continue with Facebook
