Disciplined Convex-Concave Programming¶
Discuss with fellow optimization enthusiasts.
DCCP extends CVXPY to solve nonconvex optimization problems using an organized heuristic for convex-concave programming.
Documentation contents¶
📦 Installation - Get started quickly
📖 User Guide - Get the most out of this package
⚙️ Settings - Algorithm parameters and configuration
🔗 API Reference - Detailed function documentation
💡 Examples - Real-world applications
Quick Start¶
import cvxpy as cp
import dccp
# create a nonconvex problem
x = cp.Variable(2)
y = cp.Variable(2)
problem = cp.Problem(
cp.Maximize(cp.norm(x - y, 2)),
[0 <= x, x <= 1, 0 <= y, y <= 1]
)
# solve with DCCP
result = problem.solve(method='dccp')
print(f"Optimal value: {result}")
Key Features¶
Extends CVXPY: Seamlessly integrates with existing CVXPY code
Handles Nonconvex Problems: Solves problems where DCP rules are violated
Organized Heuristic: Systematic approach using convex-concave decomposition
Multiple Restarts: Built-in support for random initialization and restarts
When to Use DCCP¶
DCCP is designed for optimization problems where:
The objective or constraints are nonconvex
You need approximate solutions to NP-hard problems
All expressions have known curvature (not “UNKNOWN”)*
* some problems with unknown curvature can still be solved using DCCP. This requires
CVXPY version 1.7.2 or later. To enable solving such problems we must also disable
DCCP checks by setting verify_dccp parameter to False in the solve method. See
Settings for details.