Inferring low-dimensional microstructure representations using convolutional neural networks

TL;DR

This paper demonstrates that convolutional neural networks can effectively generate low-dimensional representations of microstructural images, outperforming traditional correlation-based methods in capturing underlying parameters.

Contribution

The study introduces a novel approach combining CNN activations with manifold learning to improve microstructure image representations in materials informatics.

Findings

CNN-based embeddings outperform correlation-based methods

Low-dimensional embeddings accurately recover generating parameters

Method enhances statistical characterization of microstructures

Abstract

We apply recent advances in machine learning and computer vision to a central problem in materials informatics: The statistical representation of microstructural images. We use activations in a pre-trained convolutional neural network to provide a high-dimensional characterization of a set of synthetic microstructural images. Next, we use manifold learning to obtain a low-dimensional embedding of this statistical characterization. We show that the low-dimensional embedding extracts the parameters used to generate the images. According to a variety of metrics, the convolutional neural network method yields dramatically better embeddings than the analogous method derived from two-point correlations alone.