Course:
ME 539: Introduction to Scientific Machine Learning
Instructor: Prof. Ilias Bilionis
Description
This course bridges the gap between traditional engineering and modern data science. Starting from the foundations of Probability Theory, we explore how to quantify uncertainty and build predictive models using Machine Learning. From Bayesian Regression and Deep Learning to Physics-Informed Neural Networks, students learn to create and fit their own models, focusing on first principles and real-world engineering applications.
What you will learn
- Uncertainty Quantification: Represent and propagate parameter uncertainty through scientific models.
- Supervised Learning: Master regression, classification, and filtering techniques.
- Unsupervised Learning: Implement clustering, dimensionality reduction, and density estimation.
- Physics-Informed Modeling: Create new models that encode physical information and causal assumptions.
- Model Calibration: Calibrate arbitrary scientific models using experimental or simulated data.
- Scientific Computing: Master the Python ecosystem (NumPy, SciPy, PyTorch, PyMC3) for data analytics.
- Reproducible Research: Build and visualize data sets within interactive Jupyter Notebooks.
Course Logistics & Resources
Course Communication
- Piazza Discussion Forum: Our primary platform for Q&A and peer-to-peer discussion.
- Office Hours:Conducted via Zoom. The schedule is available within the course Brightspace.
- Direct Contact: Reach out to the teaching team via email for personal or administrative matters.
Required Tools
Online Textbook: Data Analytics for Scientists and Engineers.
Jupyter Notebooks: Used for all Reading, Hands-on, and Homework activities to ensure reproducibility.
Google Colab: Recommended cloud platform for running Python code without local installation.
GitHub Repository: Access all course activities and notebooks via this link.
Grading & Assessment
Grading Policy
Your final grade is based 100% on eight (8) homework assignments. These assignments are hybrid, covering:
- Theoretical Proofs: Derivations and mathematical foundations of ML models.
- Computational Tasks: Building, training, and calibrating models using real-world data.
- Submission:All work is submitted via Gradescope as a Jupyter Notebook or a scanned PDF.
Generative AI Policy
- Permitted Use: Students are encouraged to use AI for code assistance, concept clarification, and drafting content.
- Responsibility: You are responsible for the correctness of your submissions. Randomly generated content without personal understanding will result in deductions.
- Goal: Use AI to polish your work and focus on advanced scientific and mathematical questions.
Teaching Assistant:
- bwassgre@purdue.edu
- Office Hours (ET): Mon: 3:30 - 7:00 PM , Thurs: 11:00 AM - 1:00 PM , Fri: 4:00 - 6:00 PM
