| dc.description.abstract | In this project we analysed the functional role of directed simplicial structures in biological neural networks. This was inspired by previous work analysing the correlations in the activity in such networks, but a novel approach was here taken, emphasising the causal structure and dependencies in the dynamics of the networks. The motivation for such an approach was the belief that such a causal description more strongly reflect the underlying causal mechanisms of the systems, such that detecting these are more conducive to a complete understanding of the system at multiple levels. The datasets considered were computationally generated neural activity in a variety of networks with different simplicial structures. It was then tested whether the low-level functions considered were dependent on the size and completeness of the directed simplices. Additionally, we investigated whether the simplicies formed coherent functional groups with regard to the metrics chosen. As well as testing for specific functions, a model-independent approach was also tested, where we attempted to train machine learning models to predict the simplicial structures of the networks from their activity. From the results obtained, we were unable to attribute any functional properties to the simplicial structures, so the question of whether these structures serve any clear functional role remains unanswered. Further, we found a very high within-group variance with respect to the different functions considered, indicating that grouping the simplices together based on their size and completeness is not informative of the functional properties of the activity in these networks. This led to a more critical discussion regarding the methodological approach taken in this project, as well as the pursuitworthiness of such simplicial structures in future neuroscientific research. | eng |