April 3–5, 2023 • Online Course

Course Curriculum

The course will include five, half-day, virtual modules, with supporting materials. Registrants receive access to all course materials. Registrants will have access to materials and on-demand recordings until May 8, 2023.

Course Schedule

Course Modules

Module 1: Introduction to AI in Materials Science

Instructor: Raymundo Arróyave

This module will focus on providing a historical and conceptual perspective on the use of artificial intelligence (AI) and machine learning (ML) frameworks to solve materials science problems over the past decade. We will discuss the nature of the ‘materials science’ problem in the context of ML and AI-enabled science and engineering. Some examples of the application of ML/AI in materials research will be provided. Such examples include basic inference problems, generative models, active learning, and physics-informed ML, to name a few.

Learning Objectives

• Learn the basic ingredients of closed-loop design
• Understand how to perform simulations to assess modeling and algorithmic choices
• Appreciate the practicalities of developing their own autonomous platforms

Module 2: Introduction to Machine Learning and Deep Learning for Materials Science

Instructors: Ankit Agrawal, Bryce Meredig

The increasing availability of data from the first three paradigms of science (experiments, theory, and simulations), along with advances in AI and ML techniques has offered unprecedented opportunities for data-driven science and discovery, which is the fourth paradigm of science. Within the arena of AI/ML, deep learning (DL) has emerged as a game-changing technique in recent years with its ability to effectively work on raw big data, bypassing the (otherwise crucial) manual feature engineering step traditionally required for building accurate ML models. This module will introduce some of the fundamental concepts in AI/ML/DL along with illustrative examples of their application in materials science and engineering. We will discuss the unique aspects of applying AI/ML/DL to materials science specifically, and also how to be an informed "consumer" of these data-driven models.

Learning Objectives

• Become familiar with the basic components of a data-driven AI/ML/DL workflow
• Get a high-level understanding of the workings of commonly used AI/ML/DL algorithms and evaluation methodologies
• Learn how to gainfully apply AI/ML/DL on materials data for property prediction and inverse design with real-world examples
• Learn why materials science is a unique use case for AI/ML/DL, and some domain-specific challenges and considerations associated with applying these tools

Format: Live instruction with slides and Q&A

Module 3: Computer Vision with Deep Learning for Materials

Instructors: Dane Morgan, Ryan Jacobs, Benjamin Afflerbach

This module will focus on modern deep learning methods for image analysis and their applications in materials science. It will introduce convolutional neural networks and their use in classification and object detection. The first part of the module will include lectures that describe basics of model structure and training with a goal of guiding essential user choices related to hyperparameters, data augmentation, and transfer learning. The second part of the module will be hands on examples of classification and object detection from standard machine vision datasets and electron microscopy.

Learning Objectives

• Understand how new deep learning machine vision tools can be used in materials applications related to images
• Gain adequate knowledge of convolutional neural networks (CNNs) and their training to be able to utilize such models and understand their use in applications from others
• Learn to use a Jupyter notebook to fit, assess, and apply deep learning classification and object detection models to materials images

Module 4: Autonomous Research: Theory and Implementation

Instructors: Benji Maruyama, Kristofer Reyes

This tutorial describes the basic theory underpinning many autonomous closed-loop research platforms. Additionally, we describe the practical process of connecting the developed models and algorithms to machine-controllable equipment. Finally, we close with a discussion of various software tools relevant to autonomous research.

Learning Objectives

• Learn the basic ingredients of closed-loop design
• Understand how to perform simulations to assess modeling and algorithmic choices
• Appreciate the practicalities of developing their own autonomous platforms

Format: Synchronous / asynchronous lecture and hands-on, cloud-based tutorial 

Module 5: AI/ML for Materials Manufacturing: Understanding the Applications, Building Predictive Modeling, and Uncertainty Quantification

Instructors: David Blondheim Jr., Vipul Gupta, Sayan Ghosh

The first half of this module will review production examples of applications of data collection and machine learning in production manufacturing environments. With a focus on aluminum die casting, topics including data collection misclassifications, dealing with highly unbalanced data sets, data space overlap, and cost/accuracy needs of machine learning (ML) systems will be reviewed. In the second half, we will discuss probabilistic machine learning focusing on building Gaussian Process model and demonstrate its use for predictive modeling, uncertainty quantification, and sensitivity analysis. Lastly, we will discuss active learning or intelligent experimentation, which can reduce experimental cost for model building by ML-enhanced intelligent decision making for experimental planning. Jupyter notebook (with Python) will be used to demonstrate the modeling process vis simple examples. Hand-on activity: attendees will apply Prob-ML code using their formatted dataset and model results will be briefly discussed at the end of the workshop.

Learning Objectives

• Understand applications and challenges of data collection and ML within production manufacturing environments (focus on aluminum die casting)
• Complete cost analysis and understand the critical error threshold for ML implementation in manufacturing
• Build a good predictive model, interpret model output, and quantify uncertainty for additive manufacturing parameter development
• Learn how to use predictive model to intelligent design future experiments

• First Section (Blondheim): Live instruction, in-class homework activity, review
• Second Section (Gupta/Ghosh): Live instruction, hands-on activity, quiz

For More Information

For more information about this course, please contact: