Skip to content

unit05_plan_gemini.md

Latest commit

 

History

History
12 lines (9 loc) · 1016 Bytes

unit05_plan_gemini.md

File metadata and controls

12 lines (9 loc) · 1016 Bytes

Lecture Plan: Unsupervised Learning in Materials (ML-PC Unit 05)

Overview

This lecture applies the unsupervised methods learned in the mathematical foundations course directly to materials characterization, focusing on the "unlabeled data" problem in materials science.

Part 1: Applied Clustering (45 mins)

  • Case Study 1: Automated phase segmentation in EDS/EDX maps using K-Means and GMMs. Grouping pixels into distinct chemical phases without manual labels.
  • Case Study 2: Anomaly and defect detection in acoustic or thermal sensor streams during manufacturing processes.

Part 2: Applied Autoencoders (45 mins)

  • Case Study 1: Compressing high-dimensional microstructures. Using Convolutional Autoencoders (CAEs) to create a low-dimensional "materials latent space" from 3D X-ray Tomography volumes.
  • Case Study 2: Exploring the latent space. Clustering encoded latent vectors to automatically discover categories of defects or structural motifs without human supervision.