In this thesis, we investigated the general framework of Graph Neural Network for node and graph classification tasks. We studied the phenomenon of over smoothing and conducted experiments to validate them. We propose a ...