Topological Characteristics of Neural Manifolds
No Thumbnail Available
In recent years, neural population activity has been analysed by treat- ing it as a point cloud supported on a manifold whose structure gives information for the the type of computation that the network can per- form and the features it can represent. Simultaneously a data focused approach to topology, which is a fundamental property of manifolds, known as topological data analysis (TDA), has also emerged. We use a method from that toolbox called persistent homology, it essentially finds the holes of different dimensions and sizes in point clouds and helps us understand the underlying manifold. We study the topology of neural populations by creating theoretical models capable of recreating a particular manifold's topology in their activity and also analysing the topological structure of neural activity during spontaneous and stimulus induced states in mouse cortex. We find significant differences between the topological structure of neural manifolds for different stimulus conditions across the brain.
Faculteit der Sociale Wetenschappen