Intro

Course Topics

Tasks in Materials Informatics

Featureization

Material Informatics Algorithms

Closing Remarks

Intro

Why use data science?

Follow-up Studies

First 3D printable aluminum alloy

Material Informatics in the Early 2000s

Star Trek example

Material Genome Initiative

Goals of Materials Genome Initiative

Data Science Approach

Emerging Technologies in Materials Informatics

Intro

Computational screening

Where do we get data?

Material property descriptors

Support Vector Regression

Bayesian Models

Super-hard materials

Extrapolation as a classification problem

Attention-based learning approach

Epilogue

Dr Chris Farrow

Dr Chris Farrell

Get Started with What You Have

Mapping Out Your Team's Process To Identify Bottlenecks and Improvement Opportunities

Tip Number Four Beware of Pitfalls with Small Data

Nested Cross-Validation

How Is Data Recorded

Automated Image Analysis Solution

Complement Small Data with Large Domain Expertise

Adaptive Experimental Design

Intro

Why learn materials informatics?

Machine Learning vs Materials Informatics

Material Science Datasets

Data Heterogeneity

Shore Hardness Test

Uncertainty Quantification in Materials Science

Feature Engineering vs Machine Learning

Feature Engineering Example

Choosing Machine Learning Algorithms

Acknowledgements

Featureization

Generative Adversarial Networks (GANs)

Effectiveness of the algorithm

Generative Models

Constraints on the model

Matching distributions

Finding unique atoms

Are ML models available in R?

Closing Remarks

getting ready to use jupyter notebooks

cleaning data

splitting data

manually splitting data to prevent leakage

building classic models

Principles of Sustainability

Environmental Sustainability

Social Sustainability

Economic Sustainability

Materials Informatics

Why Informatics?

Systematic progression through high-dimensional design space

Bayesian optimization vs metallurgist intuition

Machine learning case studies

How do you incorporate human knowledge?

Genetic algorithms and mutation operators

Adopting data standards

Contextual reasoning & adaptation

Outro

Acknowledgments

Paradigms of science

Big Data Analytics

Examples of Forward Models

Data Mining Example

Deep Learning Models

Data Collection and Processing

Generative Adversarial Networks

Infrastructure inspection using Machine Learning

Closing Remarks

Where do we find materials data?

FAIR data principles with bad examples

Mining data from literature articles

lists of materials databases

DFT databases

non-DFT databases

benchmark datasets

crystal structure databases

file formats

14 Grand Challenges for Humanity in the 21st Century

How are Materials Discovered?

Time Required for Experiment vs. Computation

Types of Materials Data

ML Paradigms in Materials Science

Requirements for an Ideal Materials Descriptor

Graph Convolutional NNs

E(3) Equivariant GNNs

"Materials Informatics" in Industry?

Acknowledgments

Data Science for Materials Discovery

Bayesian Inverse Material Design

Generator Model

Conclusion

Introduction

Challenges

Predicting mechanical properties

Characteristic texture

How to bake in domain knowledge

Model predictions

Maximizing Likelihood of Improvement (MLI)

Design Space Quality

Can a machine learning model learn conservation laws?

Conclusion

Greg Mulholland

Limitations of Materials

Examples of Materials

Lightweight Aerospace Alloys

Canonical Measurements

Gorilla Glass

Materials Mysteries

Solar Cell

Interpretability

Product Evolved

Enthought Quick Facts

5 Stages of R&D Lab Transformation

Legacy Scientific Data Generation and Storage System Network Shared Drive

Expert-driven Data Modeling

Rapid Value from Centralized Data: Advanced Analysis

Custom Statistical Analysis

Develop Custom Machine Learning Models

Key Benefits of a Transforming Lab

Computational Methods Can Produce Informative Features for Machine Learning

Final Thoughts