Skip to content

Latest commit

 

History

History
66 lines (53 loc) · 4.04 KB

nsopt.md

File metadata and controls

66 lines (53 loc) · 4.04 KB
layout title description
maths
Teaching
Machine Learning for Materials Science

Description

Machine learning has received a lot of hype over the last decade, with techniques such as convolutional neural networks and deep learning powering a new generation of data-driven analytics. On the other hand, material science has benefited a lot from large-scale modeling & simulation through Molecular Dynamics, Density Functional Theory, and Differential Equations describing rigorous scientific laws. This course aims to provide students trainings with a convergence of the two disciplines. We will start from machine learning basics, its mathematical foundations, then move on to modern machine learning methods for material science problems and hands-on study with Python. Particularly, students will learn about how to combine the data-driven ML techniques with existing knowledge of material science to give reliable physical predictions. Various case studies will be discussed, with real-world material science applications.

Lectures

2022 Spring: Tuesday and Thursday, 3:00–4:30pm, remotely via Canvas -> Zoom

Lecture Notes (available upon request)

  1. Introduction on Machine Learning
  2. Mathematical Preliminaries
  3. TensorFlow, Fourier Analysis and Nyquist Sampling Thm
  4. PyTorch, Over/Under Fitting
  5. Convolutional Neural Network (CNN)
  6. Recurrent Neural Network (RNN)
  7. Ensemble Learning for Materials Feature Prediction
  8. GAN, ResNet and GCN
  9. Mathematical Theory and Scientific Applications
  10. Deep Learning for Partial Differential Equations
  11. Physics-Informed Machine Learning (PINN) and DeepXDE
  12. Physics Inspired Machine Learning
  13. Support Vector Machine (SVM) and kernel methods
  14. Dimension Reduction and Metric Learning for High-dimensional data
  15. Clustering Techniques and Applications to 2D Ising Model
  16. Reinforcement Learning

References

  • N. Thuerey, P. Holl, M. Mueller, P. Schnell, F. Trost, K. Um. Physics-based Deep Learning. Freely Available at physicsbaseddeeplearning.org
  • I. Goodfellow, Y. Bengio, A. Courville. Deep Learning. The MIT Press, 2016.
  • Koki Saitoh (translated by Yujie Lu). Deep Learning from Scratch (in Chinese). O’Reilly Japan, Inc.
  • Zhihua Zhou. Machine Learning (in Chinese). The Tsinghua Press, 2016.