UT - Portugal Scientific Machine Learning

Krishna Kumar

A 6-hour synchronous online course designed to equip engineers, scientists, and technical professionals with the fundamental principles and practical skills of Scientific Machine Learning (SciML) and its application in accelerating scientific computing tasks. The course focuses on three key pillars: integrating physics into neural networks (PINNs), learning solution operators for families of PDEs (Operator Learning), and leveraging automatic differentiation for end-to-end differentiable simulations. Participants will gain hands-on understanding through demonstrations using industry-standard tools like Python, PyTorch, and JAX within the Google Colab environment. The course is structured into three focused 2-hour modules, designed for live online delivery.

Course Objectives

Upon completion of this course, participants will be able to:

1. Understand the core principles of SciML, combining data-driven methods with scientific domain knowledge (PDEs/ODEs).
2. Implement and apply Physics-Informed Neural Networks (PINNs) to solve both forward (PDE solving) and inverse (parameter discovery) problems.
3. Grasp the concept of Operator Learning and implement basic DeepONets to accelerate the solution of parameterized PDEs.
4. Utilize Differentiable Programming and Automatic Differentiation (AD) to build differentiable scientific simulations for applications like gradient-based optimization and solving complex inverse problems.
5. Gain practical exposure to relevant Python libraries (PyTorch, JAX) for SciML applications.

Target Audience

This course is designed for:

• Engineers & Scientists : Professionals in various domains (Mechanical, Civil, Chemical, Environmental, Aerospace, etc.) who utilize computational modeling and simulation and seek to integrate AI/ML for enhanced performance and discovery.

• Researchers : Academics and R&D professionals looking to apply cutting-edge SciML techniques to their research problems.

• Data Scientists & ML Practitioners : Individuals with ML background interested in applying their skills to scientific and engineering domains.

• Technical Managers : Decision-makers aiming to understand the potential of SciML and accelerated computing for their teams and projects.