Explore a comprehensive tutorial on implementing a basic machine learning workflow for predicting materials properties, specifically focusing on bandgap prediction from material compositions. Learn core concepts of machine learning, including data cleaning, feature generation and engineering, model assessment, training, hyperparameter optimization, and making predictions. Follow along with a practical demonstration using the Machine Learning Lab tool, addressing common challenges in materials data and their solutions. Gain insights into decision tree structures, inputs, and outputs, as well as key distinctions in machine learning approaches. Perfect for materials science and engineering students and professionals looking to apply machine learning techniques to materials property prediction.
Introduction to a Basic Machine Learning Workflow for Predicting Materials Properties
Overview
Syllabus
A Basic Workflow for Predicting Materials Properties
Summary
An Application: Predict a Materials Property
A Basic Materials Design Workflow
Machine Learning for Pattern Matching
Key Distinction in ML
Key Distinction in ML
Model Types
Decision Trees: Structure
Decision Trees: Inputs
Decision Trees: Outputs
Summary
Machine Learning Lab Module Demo
1. Data Cleaning and Inspection
2. Feature Generation
3. Feature Engineering
4. Setup for Model Evaluation
5. Fitting and Evaluating a Default Model
6. Improving the Model by Optimizing Hyperparameters
7. Making Predictions
Questions
Taught by
nanohubtechtalks
