Evolutionary Algorithm Development for Applications in Brain Connectomics and Other Complex Systems

点击翻译按钮获取中文摘要

The brain is a computational system that performs an astronomically large number of calculations per second; however, it works in a different way than a desktop computer. The brain is made up of billions of distributed neural units that work together to perform complex functions. Unfortunately, exactly how this is done remains unknown. Brain function is what enables us to understand the world around us and solve problems; yet ironically the system - one that has enabled humans to develop complex societies and fly to the moon - remains so poorly understood by itself. The proposed research program will address this in several ways. New evolutionary algorithms (a classification of algorithms inspired by evolution) will be developed to enable the modelling of this complex computational system - the human brain. This increasingly important type of algorithm is particularly well suited for such poorly understood systems as it is less constrained and can provide novel perspectives since, by using a machine to make decisions, we eliminate many human assumptions about the problems being studied. These algorithms will be used to decipher which areas of the brain are working together, how the relationships between these areas change over time, and how the physical connections between them develop. Current modelling strategies for finding these relationships are very effective at describing most of these relationships; however, there are limitations and the popular strategies are mathematically incapable of truly modelling the underlying system. Removing these constraints is a nontrivial task that requires an unconventional solution. The careful development of these algorithms is important as it will enable us to remove many limitations on the modelling techniques to create more accurate models of the brain. With more accurate models of the brain, we can better understand how it works as a computing system, which can lead to better human-engineered systems of computation and algorithms. These models also have important clinical applications. With current modelling techniques, variations between individuals' models can be used to predict certain medical conditions, and more accurate models may allow us to diagnose and find differences more effectively. Learning how the brain develops and grows into a complex collection of wired connections can teach us about constraints on the brain, and even provide insights or explanations for why it grows the way it does. Although the focus is the human brain, the algorithms to be developed are widely applicable and will be made publicly accessible to be used by others to enable their research. Complex natural systems of information processing can be found anywhere from complex human brain networks to an ant colony working together. Understanding these systems of computation is important as it can give us new insights into different types of computation and lead to novel algorithms for solving complex problems.