- class darkmappy.solvers.PrimalDual(data, phi, psi, options={'constrained': True, 'iter': 5000, 'nu': 0, 'positivity': False, 'real': False, 'record_iters': False, 'tol': 1e-05, 'update_iter': 50})
Class which handles all primal dual optimisation paradigms.
- __init__(data, phi, psi, options={'constrained': True, 'iter': 5000, 'nu': 0, 'positivity': False, 'real': False, 'record_iters': False, 'tol': 1e-05, 'update_iter': 50})
Construct primal dual general class.
Any additional details should be here.
- Parameters
() (beta) – Measurement operator (weights for poisson noise)
() – Redundant dictionary (wavelets etc.)
() – (?)
constrained (bool) – Constrained vs unconstrained problem
- Raises
ValueError – Data vector contains NaN values.
- l1_constrained_gaussian(warm_start, sigma, beta=0.01)
Solve constrained l1 regularisation problem with Gaussian noise.
Can be instantiated from warm_start.
- Parameters
() (beta) – Data-set to be optimised over.
() – Initial solution of optimisation.
() – Noise-level present in optimisation.
() – Scaling for l1-norm threshold
- Raises
ValueError – Datavector size is 0 (empty set).
ValueError – Datavector contains NaN values.
- l1_unconstrained_gaussian(warm_start, sigma, beta)
Solve unconstrained l1 regularisation problem with Gaussian noise.
Can be instantiated from warm_start.
- Parameters
() (beta) – Data-set to be optimised over.
() – Initial solution of optimisation.
() – Noise-level present in optimisation.
() – Regularisation parameter
- Raises
ValueError – Datavector size is 0 (empty set).
ValueError – Datavector contains NaN values.
- l1_unconstrained_gaussian_jm(warm_start, sigma, beta)
Solve unconstrained l1 regularisation problem with Gaussian noise.
Can be instantiated from warm_start.
- Parameters
() (beta) – Data-set to be optimised over.
() – Initial solution of optimisation.
() – Noise-level present in optimisation.
() – Regularisation parameter
- Raises
ValueError – Datavector size is 0 (empty set).
ValueError – Datavector contains NaN values.