MS4672: Machine Learning for Materials Design

Academic Units2
Semester1
Pre-requisite(s)MS0003; MS1008
Co-requisite(s)Nil

Course Instructors

Associate Professor Kedar Hippalgaonkar

Course AIMS

This course is designed to equip you with the essential skills and practical knowledge to harness machine learning techniques for accelerating materials discovery and design. Specifically tailored for students interested in materials science, chemistry, physics, and engineering, it provides hands-on experience with core and advanced machine learning methods—including neural networks, optimization strategies, and generative modelling—to tackle real-world materials science problems. By mastering these data-driven approaches, you'll enhance your research capabilities, prepare for cutting-edge industry roles, and lay a strong foundation for future coursework or careers at the intersection of artificial intelligence and materials innovation.

Intended Learning Outcomes

By the end of this course, you (as a student) would be able to:   

  1. Apply foundational machine learning techniques to analyze and interpret materials science data.
  2. Build and evaluate predictive models such as neural networks for material property prediction.
  3. Design and optimize experiments using machine learning-driven strategies.
  4. Represent and model materials effectively using various data-driven approaches and computational tools.
  5. Utilize generative models to accelerate discovery and design of novel inorganic materials.

Course Content

  1. Review of Machine Learning Basics: Demonstrate data analysis and optimisation tools in python; Linear regression; Overfitting/regularization; Classification algorithms; Unsupervised learning: clustering/PCA.
  2. Regression Models and Artificial Neural Networks: Principles; Building models, Feature Importance for root cause analysis, SHAP Analysis; Convolutional Neural Networks for image data; Finite System Application example – Images and other materials science datasets.
  3. Design of Experiments and Optimization for experiments: Principles; Gradient-based, gradient-free approach, constrained and unconstrained optimization; Exploration vs Exploitation, Local vs Global Optimization; Demonstrate data analysis and optimization tools in python to analyze the dataset; Real-life examples.
  4. Crystal and Material Representations: Principles; Molecules, Crystalline Materials, Alloys and more; Vector-based, Graphs, SMILES and Chem/Mat-informatics; Introduction to Machine Learnt Interatomic Potentials; Materials Project, Real material examples.
  5. Generative Models: Principles; Variational Auto-Encoders (VAEs), Transformers and Diffusion Models; Demonstrate tools in python to analyze inorganic crystals using pymatgen; Real-life example.

Reading and References

  • An Introduction to Statistical Learning (with applications in Python) (James, Witten, Hastie, Tibshirani, 2013) https://www.statlearning.com/
  • Artificial Intelligence: A Modern Approach (S. Russell and P. Norvig, 2020) (4th edition, Pearson). Machine Learning for Materials Science:
  • Machine Learning in Materials Science (Keith T. Butler, Felipe Oviedo, Pieremanuele Canepa, 2022)
  • Deep Learning for Molecules and Materials (Andrew White, 2022) https://dmol.pub/ (Fully Open Access) Materials Informatics and Representations:
  • Materials Informatics: Methods, Tools and Applications (Olexandr Isayev, Alexander Tropsha, Stefano Curtarolo, eds., 2019 https://onlinelibrary.wiley.com/doi/book/10.1002/9783527802265
  • Elements of Statistical Learning (advanced reference on ML fundamentals) (Hastie, Tibshirani, Friedman, 2013) https://hastie.su.domains/ElemStatLearn/  (Free PDF available online)