Official Resources
- Source Repository: https://github.com/materialsinnovation/pymks
- Documentation: https://pymks.readthedocs.io/
- PyPI: https://pypi.org/project/pymks/
- License: Open source (MIT)
Overview
pyMKS (Materials Knowledge System) is a Python framework for materials data analytics using the Materials Knowledge System paradigm. It provides tools for microstructure quantification, localization relationships, and homogenization/linkage using 2-point statistics and MKS regression.
Scientific domain: Materials Knowledge System, microstructure quantification, 2-point statistics
Target user community: Researchers analyzing microstructure-property relationships in materials science
Theoretical Methods
- 2-point spatial correlations (statistics)
- Materials Knowledge System (MKS) regression
- Microstructure quantification
- Localization relationships
- Homogenization linkages
- Discrete Fourier transform for correlations
Capabilities (CRITICAL)
- 2-point spatial correlation calculation
- MKS regression for localization
- Microstructure generation and sampling
- PCA of microstructure statistics
- Homogenization linkages
- Discrete Fourier transform acceleration
Sources: GitHub repository, ReadTheDocs
Key Strengths
Microstructure Analysis:
- 2-point statistics
- Microstructure quantification
- PCA dimensionality reduction
- Statistical representation
MKS Framework:
- Localization (micro→local property)
- Homogenization (micro→effective property)
- Regression-based linkages
- Efficient computation via FFT
Data-Driven:
- No physics simulation needed
- Purely data-driven
- Statistical learning
- Scalable
Inputs & Outputs
- Input formats: Microstructure images, 2D/3D grids
- Output data types: Spatial correlations, property predictions, MKS coefficients
Interfaces & Ecosystem
- NumPy: Computation
- scikit-learn: ML models
- matplotlib: Visualization
- Python: Core language
Performance Characteristics
- Speed: Fast (FFT-based)
- Accuracy: Data-dependent
- System size: Microstructure grids
- Memory: Moderate
Computational Cost
- 2-point statistics: Seconds to minutes
- MKS regression: Minutes
- No DFT needed: Image/data-based
Limitations & Known Constraints
- Microstructure focus: Not for electronic structure
- Grid-based: Requires discretized microstructure
- Linear assumption: Basic MKS is linear
- Data quantity: Needs sufficient samples
Comparison with Other Codes
- vs pymatgen: pyMKS is microstructure, pymatgen is atomic
- vs matminer: pyMKS is spatial statistics, matminer is descriptors
- vs DREAM3D: pyMKS is Python, DREAM3D is C++ with GUI
- Unique strength: Materials Knowledge System with 2-point spatial correlations and MKS regression for microstructure-property linkages
Application Areas
Microstructure-Property:
- Structure-property linkages
- Homogenization models
- Localization predictions
- Process-structure-property
Materials Design:
- Microstructure optimization
- Inverse design
- Statistical learning
- High-throughput screening
Best Practices
Data:
- Use sufficient microstructure samples
- Check spatial correlation convergence
- Validate with known systems
- Use PCA for dimensionality reduction
Community and Support
- Open source (MIT)
- PyPI installable
- ReadTheDocs documentation
- Materials Innovation team
- Published in Computational Materials Science
Verification & Sources
Primary sources:
- GitHub: https://github.com/materialsinnovation/pymks
Confidence: VERIFIED
Verification status: ✅ VERIFIED
- Source code: ACCESSIBLE (GitHub)
- PyPI: AVAILABLE
- Specialized strength: Materials Knowledge System with 2-point spatial correlations and MKS regression