A computational framework of associative learning and memory formation

When a musician plays piano, she needs to recall a sequence of notes. Neuroscientists seek to understand how brain activity gives rise to a timed sequence of events, as a result of associative learning and memory formation. These behaviours are formed within neuronal networks, which consist of the building blocks of single cells and connections between cells (synapses). Most neuroscience research is focused on neural signalling that happens in individual brain nerve cells (cortical neurons). However, there are other types of activities within and above individual neurons. Within neurons, each individual neuron has a tree-like structure ending in thousands of tiny connection points (synapses forming on spines). Above any one neuron, there are populations of neurons that work in concert within a microcircuit. Both synaptic spines and neuronal populations are activated and represented as neural signals. Recent advances in experimental techniques enable observation of neural signalling at both spatial scales. At fine scale, one can measure activities of a subset of synapse/spines in one neuron. In contrast, at the coarse scale, one can simultaneously observe a subset of neurons in mutual activation. When we learn a sequence of music notes, it is likely that neuronal signals at spines and populations display similar dynamics, and that they are activated in a sequence after learning. However, the functions and mechanisms of sequential learning and memory are not well understood, as it is quite difficult to observe both types of activity simultaneously. Thus, a computational approach is necessary, as it will give us new insights into the learning process that happens on both spatial scales. 

The aim of this project is to describe learning process leveraging neuroscience data in animals and humans and developing a novel modelling framework using biological principles of neurons and synapses. The outcome of this project will provide new insights into the memory mechanisms of humans, contribute to novel neural network models for learning and memory, and develop novel methodologies for next-generation artificial intelligence. 

The supervision team has expertise in machine learning, computational neuroscience, animal and human neuroscience, brain disease, clinical neuroscience, with regularly published research papers at the very top venues in related fields.