ADMM Documentation

Start with Overview for product scope and solver coverage. Use Installation to confirm platform support and install the package. See User Guide for the modeling workflow, Examples for representative formulations, and the Python API for exact symbols and signatures.

In these docs, ADMM refers to the product/documentation name, while admm is the Python package.

Contents

• 1. Overview
• 2. Installation
  • 2.1. Supported Platforms
  • 2.2. Install ADMM Python Library
    • 2.2.1. Install from PyPI
    • 2.2.2. Install from Source
    • 2.2.3. Isolated Python Environment
    • 2.2.4. Installation Verification
    • 2.2.5. Troubleshooting
• 3. User Guide
  • 3.1. Minimal Model
  • 3.2. Supported Problem Structure
    • 3.2.1. Map This Abstract Form to Code
    • 3.2.2. Support Boundary
  • 3.3. Modeling Workflow
  • 3.4. Variables
    • 3.4.1. Choose the Shape First
    • 3.4.2. Attributes vs. Explicit Constraints
    • 3.4.3. Accessing Solution Values
    • 3.4.4. Warm Start
    • 3.4.5. Basic Operators
    • 3.4.6. Variable Shapes and Transformations
  • 3.5. Parameters
  • 3.6. Objective
    • 3.6.1. Build the Objective in Layers
    • 3.6.2. Common Objective Ingredients
  • 3.7. Constraints
    • 3.7.1. How to Choose the Direct Form
    • 3.7.2. Walkthrough: Projecting Onto a Feasible Set
  • 3.8. Solver Options
  • 3.9. Solve the Model
    • 3.9.1. Read Results in This Order
    • 3.9.2. Error Handling and Troubleshooting
    • 3.9.3. What to Do When the Solve Does Not Succeed
  • 3.10. Model Save and Reload
    • 3.10.1. What Is Saved
    • 3.10.2. What Is Not Saved
    • 3.10.3. Compression
    • 3.10.4. Example: Save, Reload, and Re-solve
  • 3.11. Supported Building Blocks
    • 3.11.1. Three Modeling Roles
    • 3.11.2. Worked Example: One Model, Three Roles
    • 3.11.3. Elementwise Atoms
    • 3.11.4. Vector and Matrix Atoms
    • 3.11.5. Structural Variable Attributes and Domain Constraints
    • 3.11.6. Compact Atom Reference
  • 3.12. Symbolic Canonicalization
    • 3.12.1. Write Clean Math First
    • 3.12.2. Walkthrough: Readable Math, Recognized Structure
    • 3.12.3. Mathematical Rewrite Examples
    • 3.12.4. Transform Examples With Auxiliary Variables
  • 3.13. User-Defined Proximal Extensions
    • 3.13.1. What a UDF Provides
    • 3.13.2. Walkthrough: The L0 Norm as a UDF
    • 3.13.3. Walkthrough: Log-Cosh Loss via grad
• 4. Examples
  • 4.1. Linear Program
  • 4.2. Quadratic Program
  • 4.3. Semidefinite Program
  • 4.4. Second-Order Cone Program
  • 4.5. Least Squares
  • 4.6. Ridge Regression
  • 4.7. Huber Regression
  • 4.8. Sparse Logistic Regression
  • 4.9. SVM with L1 Regularization
  • 4.10. Quantile Regression
  • 4.11. Robust PCA
  • 4.12. Sparse Inverse Covariance Selection
  • 4.13. Entropy Maximization
  • 4.14. Fault Detection
  • 4.15. Water Filling
  • 4.16. Convolutional Image Deblurring
  • 4.17. Portfolio Optimization
  • 4.18. L0 Norm
  • 4.19. L0 Ball Indicator
  • 4.20. L1/2 Quasi-Norm
  • 4.21. Group Sparsity
  • 4.22. Matrix Rank Function
  • 4.23. Rank-r Indicator
  • 4.24. The Unit-Sphere Indicator
  • 4.25. The Stiefel-Manifold Indicator
  • 4.26. The Simplex Indicator
  • 4.27. The Binary Indicator
  • 4.28. L0-Regularized Regression
  • 4.29. Log-Cosh Robust Regression
  • 4.30. Cauchy Loss Robust Regression
  • 4.31. Smooth Quantile Regression
  • 4.32. Wing Loss for Precise Regression
  • 4.33. Smooth Total Variation Denoising
  • 4.34. Gamma Regression (GLM with Log Link)
• 5. Python API
  • 5.1. Global functions
    • 5.1.1. Functions
  • 5.2. ADMMError
    • ADMMError
  • 5.3. OptionConstClass
    • OptionConstClass
  • 5.4. TensorLike
    • TensorLike
  • 5.5. Constant
    • Constant
  • 5.6. Var
    • Var
  • 5.7. Param
    • Param
  • 5.8. Constr
    • Constr
  • 5.9. TuningContext
    • TuningContext
  • 5.10. UDFBase
    • UDFBase
  • 5.11. Model
    • Model
  • 5.12. ArrayLike
    • ArrayLike
• 6. Citing ADMM
• 7. Contact