Accelerated Discovery to Knowledge using Machine Learning
summary
A materials ML workflow that ships a public MXene database and working models: a 23,000 MXene entries DFT catalog plus GW-calibrated regressors that predict band gaps and VBM/CBM in minutes with uncertainty estimates, a classifier that separates metals vs semiconductors at ~99% ROC accuracy, a lattice thermal conductivity model using four physics-derived features (including Grüneisen and phonon inputs) that outperforms Slack with error 0.21, and a new guided patchwork kriging algorithm and a prototype GUI for hardness prediction.
AI-written summary · about
chapters (23)
- 0:00Intro
- 1:05Machine Learning The goal of machine learning is to develop methods that can automatically de other outcomes of interest Kevin P. Murphy
- 2:48Materials Genome Initiative
- 4:53Materials Databases • Multitude of databases of materials host more than millions of compounds calculated properties
- 7:32A Functional Materials Database for MXene
- 9:40Feature Selection (X)
- 12:53Semiconductor-Metal (Classification)
- 14:51Band Gap Prediction
- 15:59Physical Significance ML model does not use PBE band gap of MXene for prediction
- 17:19Band Edge Alignment
- 18:56High-throughput Materials Design
- 20:10Property Map
- 22:01Prediction Model . With four proposed descriptors
- 23:44Relating Electronic to Thermal Transport
- 25:11Independent Models
- 25:35Linking Descriptors
- 27:13Linking Prediction Model Prediction model for using electronic transport derived and common bridging descriptors
- 28:20Learning from Machine
- 29:22Structural Materials
- 30:11Image processing Database: 110 microstructures and compositions
- 31:26Approach II: 2-point correlations
- 31:47Vickers hardness prediction
- 32:36Conclusion
related talks
28mHow Is Python Transforming Materials Modeling with Machine Learning?
24mMatscholar: The search engine for materials science researchers
31mML methods to accelerate the drug discovery cycle
11mImproving Drug Discovery with Machine Learning...
19mQuantum Computing: Hype, Reality and Business Relevance for Materials Discovery
26mMLbase: A Distributed Machine Learning System
27mAI-ML and space exploration at NASA
57mStartup Showcase