27-737   Data Analytics and Machine Learning for Materials Science

Location: Pittsburgh

Units: 12

Semester Offered: Spring

Course description

Materials Science and Engineering has traditionally been taught by emphasizing the development and application of technology. This course will present an alternative approach that combines data analytics and machine learning with material fundamentals (i.e. materials informatics).

Learning objectives

The main educational goal is to enable students to use data analytics and machine learning techniques in their own research.

Content details

The topics will include:

  • data collection, pre-processing, analytics, and visualization
  • unsupervised learning techniques, e.g., PCA, tSNE, DBSCAN, and k-means-clustering
  • supervised machine learning techniques, e.g., XGBoost, kNN, and random forest
  • deep learning techniques, e.g., (convolutional) neural networks, learning with memory, and transfer learning
  • the CPU and single and multiple GPU versions of these techniques

Finally, students will get experience working with inverse design techniques, e.g., Bayesian optimization and particle swarm optimization.

Prerequisites

The course does not have prerequisite courses. A background in statistics, linear algebra, and Python coding are helpful but not required.