Microstructure Quantification Through Spatial Statistics

An interactive presentation of spatial statistics and PCA

Posted by Ahmet Cecen on August 11, 2016

Recently by the same author:

Regular 2pt Statistics PCA on Half Vortex

Posted by Ahmet Cecen on March 9, 2017

Projector Visualization

Ahmet Cecen

Data Scientist / Materials Informatics

2-Point Statistics

  • 2-Point Statistics can capture inherent structural organization in data irrespective of positioning, frame, window size or origin.

  • 2-Point Statistics can highlight inherent lenght scales in the structure.

More about 2-Point Statistics.

Principal Component Analysis

PCA is a dimensionality reduction technique that extracts orthogonal features that reflect the greatest variance in data.

More about PCA.

Interactive Exploration

Example Random Dataset

Exploring the Microstructure Space

In case of user overload, you can also access the following app at:

You can also find some quick picked example patterns here

Function Estimation

We can use the above representation to fit functions of the form:

Property = f(PC1,PC2,…).

Reconstructions

We can then explore the reverse problem of how the structure would look like given a desired property value.