CSE 8803 / ME 8883 - Materials Informatics Course - Fall 2016

Discoverability
Visible
Join Policy
Invite Only
Created
20 Aug 2016

Nanocrystals: Blog Post 7 - PCA

Extracting morphological changes in nanocrystals using in situ liquid cell microscopy

PCA

After constructing two-point statistics plots for all the microstructures, these data were built into an array. The first step for PCA is to mean-center the data, so the mean of the array was calculated. This mean was then subracted from each microstructure of the arrray. PyMKS was then used to perform the SVD calculations and provide the basis vectors as well as the pc scores for every microstructure. Below is the scree plot showing how much of the variance is captured by first few pc scores.

scree_plot.JPG

Upon examining the scree plot, we are quite satisfied with retaining only the first three pc scores since 99% of the variance is captured in these first three pc score alone (the pc scores capture 96%, 2.46%, and 0.65% respectively). To show that this truncation is acceptable, below we have plotted the two-point statistics for a microstructure alongside a one-term, two-term, and three-term reconstruction of the two point statistics using the basis vectors and the pc scores.

1_pc.png

2_pc.png

3_pc.png

We now have the first two pc scores for all the microstructures and are ready to build our model.

Created on , Last modified on