Machine Learning Fundamentals for Microstructural Analysis

Hosts: Prof. Katayun Barmak (Columbia U.), Prof. Jeffrey Rickman (Lehigh U.), and Matthew Patrick (Columbia U.)

The development of a Materials Genome Initiative (MGI) workforce depends critically on the education of the next generation of scientists and engineers, particularly in modern data science techniques. Unfortunately, when machine learning tools are introduced to beginning students, they are often presented as “black-box” software modules employed without understanding of their statistical underpinning.

This interdisciplinary skills bootcamp, geared toward beginners, seeks to address this shortcoming by providing attendees with both the skills and the requisite understanding of a spectrum of data science tools. As a concrete illustration of the power and utility of these tools, we will take as an example the application of neural networks to problems in microstructural image analysis.

To express interest, please contact:

Jeffrey Rickman: [email protected]

Matthew Patrick: [email protected]

Tentative Program

Day 1, Monday, June 2

1. Introduction
- Overview
- Practical information for an Interactive Workshop  - Jupyter Notebooks, Google Colaboratory
- Description of Software

2. Data Interpretation
- Data Wrangling
- Clustering Methods, K-means clustering  (Example – classifying materials by properties)
- Principal Component Analysis – dimensional reduction
- Kriging and Interpolation
(Canonical Correlation Analysis)

_______________________________

Day 2, Tuesday, June 3

3. Probability and Statistics for Microstructural Interrogation
- Probability Distributions and Density Functions
- Moments
- Analysis of Grain-Size Distributions
- Correlation Analysis
- Tessellations
- Hypothesis Testing
- Bayesian Methods

4. Simulation Methodology for Microstructural Evolution
- Nucleation and Growth – Voronoi Models
- Coarsening – Monte Carlo method

5. Optimization
- Genetic Algorithms

_______________________________

Day 3, Wednesday, June 4

6. Classification and Regression
- Decision Trees
- Random Forests
- Support Vector Machine

7. Neural Networks for Microstructural Interrogation
- Perceptron Models
- Deep Learning
-  Backpropagation
- Errors
- Generic Examples     
- Introduction to U-Net
- Experimental Imaging of Materials Microstructures
- Microstructural Case Study

_______________________________

Day 4, Thursday, June 6

8. Practice Time / Work with Your Own Data

_______________________________

Day 5 (Half day), Friday, June 7

9. Conclusions

This boot camp was financially supported by the US National Science Foundation (NSF) under the DMREF program grant numbers DMS-2118206 and DMS-2118197.