Settings¶

Solve Method Parameters¶

The DCCP solver accepts various parameters to control the algorithm behavior:

Name	Type	Description	Allowed values	Default value
`max_iter`	`int`	Maximum number of iterations in the CCP algorithm	\((1, \infty)\)	100
`max_iter_damp`	`int`	Maximum number of damping iterations when convergence fails	\((1, \infty)\)	10
`tau_ini`	`float`	Initial value for tau parameter (trades off constraints vs objective)	\((0, \infty)\)	0.005
`mu`	`float`	Rate at which tau increases during the algorithm	\((1, \infty)\)	1.2
`tau_max`	`float`	Upper bound for tau parameter	\((0, \infty)\)	1e8
`k_ini`	`int`	Number of random projections for variable initialization	\((1, \infty)\)	1
`k_ccp`	`int`	Number of random restarts for the CCP algorithm	\((1, \infty)\)	1
`max_slack`	`float`	Maximum slack variable value for convergence	\((0, \infty)\)	1e-3
`ep`	`float`	Convergence tolerance for objective value changes	\((0, \infty)\)	1e-5
`std`	`float`	Standard deviation for random variable initialization	\((0, \infty)\)	10.0
`seed`	`int \| None`	Random seed for reproducible results	\(\mathbb{Z} \cup \{\text{None}\}\)	None
`verify_dccp`	`bool`	Whether to verify DCCP compliance before solving. Enables solving problems with `UNKNOWN` curvature	\(\{0, 1\}\)	1

Parameter Usage¶

You can pass these parameters to the solve method like so:

import cvxpy as cp
import dccp

# create your problem
x = cp.Variable(2)
problem = cp.Problem(cp.Maximize(cp.norm(x, 2)), [cp.norm(x, 1) <= 1])

# solve with custom parameters
result = problem.solve(
    method='dccp',
    max_iter=200,
    tau_ini=0.01,
    k_ccp=5,
    seed=42
)