SciANN: Un framework per la risoluzione di PDE attraverso le Physics-Informed Neural Networks

Machine Learning is a branch of Artificial Intelligence which focuses on learning automatically from data without being explicitly programmed. In recent years this field has exhibited an impressive development due to the increase of computer's capacity and processing power. In this thesis we focus on artificial neural networks, a model inspired by the human brain and the building block of deep learning. In particular we present "Physics-Informed Neural Networks" which are a type of neural network specifically designed to resolve supervised learning problems involving systems whose behavior is described by general nonlinear partial differential equations. In the end we conclude presenting with examples SciANN, a Python package which uses Keras and Tensorflow as backends to implement physics-informed neural networks for scientific computations, solution and discovery of partial differential equations.