Machine Learning in Science

URL study guide

Description

• Basics of physical modeling, scaling
• Basic of machine learning and neural networks, deep networks
• Stochastic gradient descent, backpropagation
• Introduction to symmetry in physics
• Injecting symmetries in neural networks
• Convolutional neural networks
• Symmetric networks

Objectives

1. explain basic conceptual and design aspects of modern machine learning techniques - from linear and non-linear regression to deep learning - to perform data-driven modeling of elementary physics problems.
2. apply machine learning methodologies to translate suitable (small-scale) datasets into predictive models of physical systems in a supervised learning context.
3. implement basic networks and training procedures leveraging on the backpropagation algorithm, using Python.
4. link with the fundamental physical concept of symmetry and integrate it in the modeling process to simplify elementary machine learning problems.
5. analyze network design choices, estimating quantitatively network performance and results.
6. use state-of-the-art computational tools for machine learning from the python ecosystem

Method of Assessment