Machine Learning Applications in Physics

PHY5006 / PHYG004 · Sogang University · 2026 Spring

Course Information

Instructor Young Woo Choi
Course PHY5006 (Graduate) / PHYG004 (Undergraduate)
Schedule Wed, Fri 09:00 – 10:15
Office F303C
Office Hours Mon, Tue, Thu 10:00 – 12:00 (by appointment)
Email ywchoi02@sogang.ac.kr

Lectures

# Topic Materials
01
Introduction to ML in Physics
Physics vs ML modeling / JAX basics / Autodiff
Slides Notebook
02
Why Machine Learning in Physics?
Approximation / Inference / Surrogate modeling
Slides Notebook
03
Neural Networks Basics
Linear models / MLP / Activation functions
Slides Notebook
04
Linear Models
Linear regression / Classification / Regularization
Notebook
05
Multilayer Perceptrons
MLP implementation / Training loop / Optimization
Notebook
06
Convolutional Neural Networks
Convolution / Pooling / LeNet / Ising phase classification
Notebook
07
FNN/CNN Applications in Physics
Ising classification / Galaxy morphology / AI-assisted coding
Handout Solution
08
Automatic Differentiation & Physics-Informed Neural Networks
Autodiff / JAX grad / PINN / Schrödinger equation
Lecture Handout Solution
09
microgpt: a GPT in 200 lines
Transformer / Autograd / Attention / Adam / Karpathy microgpt
Lecture Slides Notebook Handout Solution
11
Graph Neural Networks for Molecules
Graphs / MPNN / SchNet / Tetris / Chirality
Notebook
12
Equivariant Networks with e3nn-jax
Equivariance / Spherical harmonics / Irreps / Tensor product / TFN
Notebook Handout KO Handout EN
13
Production-Level Machine Learning Potentials
Behler-Parrinello / SchNet / NequIP / MACE / Allegro / Foundation potentials
Notebook
14
Generative Models I — Variational Autoencoders (VAE)
Generative models / VAE / Latent space / ELBO / β-VAE
Notebook

Important Dates

Week Event
Week 7 Individual Presentations (3-minute oral)
Week 8 Midterm week — no exam
Week 15 Term Project Presentations
Week 16 Final exam week — no exam

Projects & Presentations

Individual Presentation

Week 7 · 30% of grade

  • 3-minute oral presentation, with or without slides
  • Propose an idea for applying ML to a physics problem
  • Include motivation, methods, and expected outcomes
  • Implementation is not required

Focus on:

  • What is the physics question?
  • Why ML?
  • Inputs / outputs
  • Expected gain over traditional methods

Term Project

Week 15 · 50% of grade

  • Team-based, up to 3 students per team
  • Apply machine learning to a physics problem
  • Reproduce results from research papers, or conduct related exploratory studies
  • Present your work in a team presentation

Grading

  • Individual presentation: 30%
  • Term project: 50%
  • Participation: 20%