ECE/SIOC 228 Machine Learning for Physical Applications

Instructor

• Yuanyuan Shi, yyshi@eng.ucsd.edu

• Office hours: Wednesday 6:30 - 7:30 pm

Teaching Assistants

• Yuexin Bian

• Tz-Ying Wu

Course Time and Location

Monday/Wednesday, 5-6:20 pm, Pepper Canyon Hall 109

Schedule

Part 1: Machine Learning Fundamentals

• Week 1: Course Logistics; supervised learning setup; linear models for regression 

• Week 2: Linear models for classification; Feature selection: ridge regression, Lasso; Bias and Variance Tradeoffs

• Week 3: Neural network basics; computational graph and backpropagation; optimization and regularization

• Week 4: Temporal data modeling: RNN & LSTM/GRU; Attention & Transformers 

• Week 5: Spatial data modeling: CNN // Graph data modeling: graph neural networks

• Week 6: Graph neural networks; summary of the first part of the course; 

Part 2: Incorporate Physics/Prior Knowledge into Learning 

• Week 7: Review of ODE and PDE; Physics-informed machine learning; Neural operators 

• Week 8: Neural ODEs; OptNet; Equivariant Neural Networks;  

• Week 9: Applications to Control - model-based RL and model predictive control

• Week 10: Final project presentation (poster session or optional in-class presentation, TBD)