Review of applications of graph theory and network neuroscience in the development of artificial neural networks

One of the important subfields of computational neuroscience is network neuroscience, which widely uses machine learning techniques, these include artificial neural networks. The latter ones owe a lot to neuroscience in the history of their development. Artificial neural networks (ANN) are often represented in the form of a graph – a visual representation of connections between the elements of the system, which form a network. However, this does not mean that an ANN was designed based on graph theory or with consideration of the special properties of complex networks. These representations are often called computational graphs and are usually just an implementation of the network architecture without understanding the impact of graph structure on its performance (You, 2020). Many researchers suggest that the deliberate use of graph theory while creating artificial networks allows them to improve their prediction performance. This improvement can be applied to a variety of deep learning tasks, such as classification, searching, chemical reaction prediction, knowledge graphs, machine translation, and many more (Liu & Zhou, 2020). It also turns out that the networks with the best performance have graph properties very similar to biological nervous systems (You, 2020). In this talk, I will present the most important concepts in currently developing approaches to creating artificial neural networks: Graph Neural Networks and modular neural networks.