examples

CVXPYlayers Examples

This directory contains examples demonstrating how to use cvxpylayers across different frameworks and application domains.

Getting Started

Start with the basic examples for your framework of choice:

Framework	Quick Start	Tutorial
PyTorch	torch/torch_example.py	torch/tutorial.ipynb
JAX	jax/jax_example.py	jax/tutorial.ipynb
MLX	mlx/mlx_example.py	-

Examples by Domain

Control Systems

Example	Description
torch/lqr.ipynb	Linear Quadratic Regulator - learn optimal value function parameters
torch/constrained_lqr.ipynb	LQR with control input bounds and state constraints
torch/vehicle.ipynb	Autonomous vehicle path planning and trajectory optimization
torch/constrained_mpc.ipynb	Model Predictive Control with learned cost-to-go
torch/convex_approximate_dynamic_programming.ipynb	Dynamic programming with convex approximation

Finance & Portfolio Optimization

Example	Description
torch/markowitz_tuning.ipynb	Portfolio optimization with dynamic rebalancing
torch/Portfolio optimization with vix.ipynb	Portfolio optimization incorporating VIX volatility

Machine Learning

Example	Description
torch/ReLU Layers.ipynb	Replace ReLU activations with differentiable optimization layers
torch/monotonic_output_regression.ipynb	Learning monotonic input-output relationships
torch/signal_denoising.ipynb	Signal/image denoising with learned parameters
torch/data_poisoning_attack.ipynb	Adversarial data poisoning attack on logistic regression

Resource Allocation

Example	Description
torch/resource_allocation.ipynb	Water/resource distribution optimization
torch/supply_chain.ipynb	Supply chain network flow optimization

Physics & Engineering

Example	Description
torch/optimizing_stiffness_constants.ipynb	Optimal design - spring stiffness coefficients

Related Papers

Several examples accompany published research:

• Learning Convex Optimization Control Policies (COCP): lqr.ipynb, constrained_lqr.ipynb, vehicle.ipynb, supply_chain.ipynb, markowitz_tuning.ipynb
• Learning Convex Optimization Models: monotonic_output_regression.ipynb, signal_denoising.ipynb, constrained_mpc.ipynb
• Differentiable Convex Optimization Layers (NeurIPS 2019): data_poisoning_attack.ipynb

Dependencies

Most examples require:

• cvxpylayers with the appropriate framework (pip install cvxpylayers[torch], [jax], or [mlx])
• matplotlib for visualization
• numpy, scipy for numerical operations

Some notebooks have additional dependencies (e.g., networkx, seaborn, PIL).