ECE/SIOC 228 Machine Learning for Physical Applications

Instructor

• Yuanyuan Shi, Assistant Professor, yyshi@eng.ucsd.edu

• Office hours: Thursday 4-5 pm [Jacobs Hall Room 4401]

Teaching Assistants

• Tawaana Homavazir, Office Hours: 2:00 pm - 3:00 pm Monday 

• Rishabh Jangir, Office Hours: 3:30 pm - 4:30 pm Tuesday 

• Srinivas Rao Daru, Office Hours: 3:30 pm - 4:30 pm Wednesday 

• Rohin Garg, Office Hours: 9:30 am - 10:30 am Thursday 

Course Time and Location

Tuesday/Thursday, 2-3:20 pm, Warren Lecture Hall 2001

Course Structure

Part 0: Introduction & Fundamentals

• Lecture 1: Introduction and Course Logistics [Lecture Note]

• Lecture 2: Supervised Learning Setup [Lecture Note]

• Lecture 3: Neural Networks and Backpropagation [Lecture Note]

• Lecture 4: Optimization & Regularization in Deep Learning [Lecture Note]

Part 1: Modeling and Learning in Physical Systems

• Lecture 5: Convolutional Neural Networks  [Lecture Note]

• Lecture 6-7: Recurrent Neural Networks [Lecture Note]

• Lecture 8-9: Graph Neural Networks [Lecture Note]

• Lecture 10: Neural Operators (Guest Lecture: Zongyi Li) [Lecture Note]

• Lecture 11: Physics Informed Neural Networks [Lecture Note]

• Review of Modeling and Learning [Lecture Note]

Part 2: Decision Making in Physical Systems

• Lecture 12-13: Markov Decision Process [Lecture Note]

• Lecture 14: Deep Reinforcement Learning I: Policy Gradient [Lecture Note]

• Lecture 15: Deep Reinforcement Learning II: Policy Gradient with Variance Reduction [Lecture Note]

• Lecture 18: Deep Reinforcement Learning III: Actor-Critic Method [Lecture Note]

• Lecture 19: Deep Reinforcement Learning IV: Value Function Methods and DQN [Lecture Note]

Part 3: Guest Lectures (ML for various physical applications) and Projects

• Lecture 16: Machine Learning for Power System: Route to Sustainability (Dr. Yize Chen) [Lecture Note ]

• Lecture 17: Machine Learning for Robotics (Dr. Guanya Shi) [Lecture Note]

• Lecture 20: Final Project Poster Presentation

Textbook and References

There is no single text that provides adequate coverage of the concepts and the background material in to serve the needs of this course. There is no required textbook for the course. We will use my lecture notes and the following other references/lecture notes for material, each of which has a part of the relevant material for the course. 

Modeling and Learning in Physical Systems

Deep Learning, by Ian Goodfellow and Yoshua Bengio, and Aaron Courville

CS230 Deep Learning, Stanford, by Andrew Ng

Understanding Convolutions on Graphs, by Ameya Daigavane, Balaraman Ravindran, and Gaurav Aggarwal

Physics-Informed Neural Networks: A Deep Learning Framework for Solving Forward and Inverse Problems Involving Nonlinear Partial Differential Equations, by M. Raissi, P. Perdikaris, and G.E. Karniadakis

Decision Making in Physical Systems

An Introduction to Reinforcement Learning,  by Richard Sutton

Reinforcement Learning,  by David Silver

Deep Reinforcement Learning, by Sergey Levine