Register Now
Lead Instructor(s)
Date(s)
Jul 29 - Aug 01, 2024
Registration Deadline
Location
Live Virtual
Course Length
4 Days
Course Fee
$3,600
CEUs
2.2
Sign-up for Course Updates

Material informatics is transforming the way materials are discovered, understood, developed, selected, and used. In this condensed course, you will engage in interactive lectures, clinics, and labs designed to help you learn, design, and apply modern material informatics tools and large-scale multiscale modeling—with the ultimate goal of helping you to speed up your design process and implement cost effective rapid discovery and prototyping in your organization. 

This course may be taken individually or as part of the Professional Certificate Program in Design & Manufacturing or the Professional Certificate Program in Machine Learning & Artificial Intelligence.

Course Overview


Artificial intelligence is changing the paradigm for many industries, and materials-focused commerce is no exception, where tremendous opportunities lie ahead. With the success of effective and generalizable deep learning tools, the materials industry is primed to take advantage of unprecedented breakthroughs, leveraging materials modeling, analysis, and design toward a more efficient, less costly, and more versatile response to market demands and opportunities, through materiomics. With data available from autonomous experimentation, large databases like the Materials Project within the Materials Genome initiative, or synthetic data, there exist many opportunities to accelerate and expand your materials design platform.

Today, practicing engineers are expected to have both domain knowledge and a solid understanding of modern machine learning tools. This course will teach all the fundamentals necessary for you to reach the next milestone in practicing materiomics, by navigating the complex world of AI. Participants will learn fundamentals and techniques to develop and deploy machine learning in materials development and gain a first-hand understanding of state-of-the-art tools for varied applications ranging from data mining to inverse design.

What's Next in Materials Science?

Computational modeling and simulation have been instrumental in advancing materials design for decades, particularly since the introduction of the finite element analysis methods (FEA/M), computational fluid dynamics, advanced imaging, and more recently, molecular modeling at the chemical level. Since the emergence of machine learning (ML) and especially deep learning, the materials field is primed for yet another revolution in which forward and inverse problems can be approached from a unique and complementary perspective. Together with advanced manufacturing, the integration of model, theory, and simulation-based design has the potential to unlock new opportunities across industries as AI systems learn to read and generate complex fingerprints of a material’s process-structure-property relationships. This opens direct pathways to understand, mine, and generate out-of-the-box solutions that address critical needs such as low-carbon design, sustainability goals, or enhanced functionality.

MIT has pioneered research and teaching of computational methods for materials science since the 1970s, and now plays a leading role in the data science revolution. Indeed, recent years have seen the development of a solid foundation of peer-reviewed work (publications, software, patents, and startups) at the nexus of materials design and deep learning. It is becoming an important frontier in engineering problem solving across scales—from molecular to structural. MIT is leading this development through research, IP creation, and startups. In this course, you will get first-hand exposure to this important area of technology in a stimulating and engaging environment.

Machine Learning for Materials Informatics Course Details

Fueled by enormous advances in machine intelligence, the field of material informatics is helping today’s professionals cut down the development time of new materials and processes from decades to weeks and months, and achieve superior sustainability goals. In this course you will fully learn how to incorporate these new technologies and methods into your own material design processes in order to capitalize on recent AI breakthroughs, such as language models (e.g. GPT-3, BERT), DNA and protein models (e.g. AlphaFold), graph neural networks for molecular to macroscale structures, and a variety of tools from computer vision, specifically for the analysis, design and modeling of materials.

Every day, hundreds of machine learning papers are published, making it challenging to keep track of the most important advances and methods. The instructor, Professor Buehler, will masterfully break down this complex field into easy-to-digest concepts, to offer you direct access to leverage the new tools for your problem space, and to develop the skill to judge and assess the best tools for the job. Alongside peers from around the world, you will engage in interactive lectures and hands-on coding clinics and labs delivered in a live virtual format. These activities are designed to help you learn, design, and apply modern material informatics tools—specifically artificial intelligence and machine learning—including neural interatomic potentials, large-scale multiscale modeling to improve the speed, efficiency, and cost effectiveness of your discovery, prototyping, and development processes. You will learn how modern computational tools enable us achieve almost any desirable accuracy in multiscale material discovery, connecting quantum to the macro-world. 

Specifically, topics covered include:

  • Modern and cutting-edge machine learning tools, especially focused on deep learning (includes: convolutional neural nets, adversarial methods, graph neural nets, autoencoders, transformer models; including neural molecular dynamics)
  • Working across data modalities: Analysis of images, voxel data, dynamical data, and graphs, as well as language and symbolic methods and hybrid approaches ; features in-depth discussion of materiomic databases, synthetic datasets, and data collection in materials development
  • Visualization and data analysis methods, including statistical methods, cluster analysis, graphic rendering, virtual reality; as well as interpretable machine learning

The course curriculum is grounded in relevant examples and case studies from a variety of fields and at distinct scales (molecular to macro), including structural materials, additive manufacturing, nanotechnology, healthcare, and biomedical engineering. Participants will gain access to codes and templates to build their own AI platforms.

NOTE: This program is designed to complement the on-campus course Predictive Multiscale Materials Design, but can also be completed independently.

Certificate of Completion from MIT Professional Education

Machine Learning for Material Informatics cert image
Learning Outcomes
  • Explore the cutting-edge of modern material informatics tools, including machine learning, data analysis and visualization, and molecular/multiscale modeling
  • Learn how to fine-tune general-purpose models for materials applications
  • Learn how to work with small, sparse, or low-quality datasets and build predictive models 
  • Deepen your knowledge of the frontiers of data-driven material analysis and ready-to-deploy code solutions
  • Master computational methods and codes for building better materials, such as language models, protein models, and graph neural networks, and how to build and use your own custom datasets
  • Learn how to identify the most effective tool for solving your specific challenge, and gain an overview across the most promising neural network architectures and their most suitable application areas, challenges and potentials; including specific code examples that will be discussed in detail
  • Solve inverse design problems using AI
  • Enhance the speed, efficiency, and cost effectiveness of your materials design and production processes through next-generation molecular modeling
  • Monetize your existing data and develop an actionable vision for incorporating material informatics into your organization’s current strategies 

PROGRAM OUTLINE

NOTE: All times Eastern Daylight Time (UTC-4:00). A few introductory lecture videos will be posted ahead of the course.

Day One

9am-noon: Foundations in material informatics (data science and basic concepts of machine learning; multiscale modeling; datasets, experimental methods for data collection)

1-2pm: Clinic #1: Convolutional neural network (classifier, regression, and peeking inside via interpretable methods)  

2-4pm: Digging deeper: Deep neural nets, loss functions, Stochastic optimization methods (e.g., stochastic gradient descent and variants), Regularization

4-5pm: Clinic #2: Material failure analysis

5-7pm: Interactive virtual networking reception (get to know peers, the instructor, and make connections)

Day Two

9-10am: Hands-on introduction to PyTorch (example application to fine-tuning a BERT NLP model applied to proteins)

10-11am: Hands-on introduction to TensorFlow (example application to developing an adversarial neural network)

11am-noon: Practical guide to tensor algebra and other important math concepts needed

1-2pm: Ethics, bias and sustainability in material informatics

2-3:30pm: Data, data, everywhere…De novo dataset construction (imaging lab) and application to build a deep neural network (covers computer vision tools, live imaging using depth camera

3:30-5pm: Introduction to graph neural networks (applications to molecular systems, truss systems, alloys, proteins, and healthcare; graph transformers)

Day Three

9-10:30am: Transforming AI and healthcare with attention (AlphaFold and applications to protein design, synthesis)

10:30am-noon: Deepening the understanding of language models applied to materials (pre-training and fine-tuning); BERT and GPT-3-like (applications of large language models to materials problems; category theory; time-dependent material phenomena)

1-2pm: Clinic #3: Transformer models for inverse materials design (develop multiscale transformer model from scratch)

2-3pm: Adversarial neural networks and applications to materials design (manufacturing, inverse problem, characterization)

4-5pm: Case study: Image segmentation in microscopy, medical imaging, and analysis

Day Four

9-10am: Autoencoders (vision, graphs, NLP, proteins)

10-11am: Clinic #4: To fail or not to fail: Buckling modeling (time-dependent phenomena)

11am-noon: Concluding discussion

Noon-12:30pm: Graduation ceremony and certificates

Post-course: Participants will have the opportunity to ask additional questions at two office hour sessions, which will be offered one and two weeks after the conclusion of the course.

 

Links and Resources

Video/Audio:

News/Articles:

Who Should Attend
  • Lead scientists or engineers who work in fields that require advanced materials design, development, or manufacturing skills.
  • Software engineers or data scientists who want to leverage data to design and produce better materials. 
  • Technology outreach directors, technology scouts, IP/patent professionals, or consultants who need to stay on the cutting-edge of intelligent material design.
  • Sustainability directors who are looking for environmentally friendly alternatives to current materials. 
  • Technical leaders or business intelligence managers/directors who need to make informed decisions related to material design strategies and investments. 
  • Entrepreneurs, founders, investors, venture capitalists, futurists, and visionaries looking to stay abreast of new opportunities in material design. 
  • Creatives and science communicators/marketers who need to understand the technologies and trends driving next-generation smart materials.
  • Policymakers/influencers who want an overview of the challenges and opportunities in materials design across industries. 

REQUIREMENTS

A computer with cloud computing access is required.

 

Brochure
Download the Course Brochure
Machine Learning for Materials Informatics - Brochure Image