TopologiCal ApproacH to Artificial NeurAL NetworKS

The project TopologiCal ApproacH to Artificial NeurAL NetworKS (CHALKS) will focus on the study of Artificial Neural Networks architecture based on computational topology approaches and sheaf theory. More specifically, towards Explainability, we plan to study the data representations through the different hidden layers of ANNs and track the topological complexity evolution. Then, we want to characterize the needed architecture based on the topological complexity.