CUQIpy 1.2.0.post0.dev305__py3-none-any.whl → 1.2.0.post0.dev333__py3-none-any.whl

{CUQIpy-1.2.0.post0.dev305.dist-info → CUQIpy-1.2.0.post0.dev333.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: CUQIpy
-Version: 1.2.0.post0.dev305
+Version: 1.2.0.post0.dev333
 Summary: Computational Uncertainty Quantification for Inverse problems in Python
 Maintainer-email: "Nicolai A. B. Riis" <nabr@dtu.dk>, "Jakob S. Jørgensen" <jakj@dtu.dk>, "Amal M. Alghamdi" <amaal@dtu.dk>, Chao Zhang <chaz@dtu.dk>
 License: Apache License

{CUQIpy-1.2.0.post0.dev305.dist-info → CUQIpy-1.2.0.post0.dev333.dist-info}/RECORD

@@ -1,6 +1,6 @@
 cuqi/__init__.py,sha256=LsGilhl-hBLEn6Glt8S_l0OJzAA1sKit_rui8h-D-p0,488
 cuqi/_messages.py,sha256=fzEBrZT2kbmfecBBPm7spVu7yHdxGARQB4QzXhJbCJ0,415
-cuqi/_version.py,sha256=CQLIC5_dwm8tSsiUeJmyVr-z-B8P5Oay1-7eVxbDQkk,510
+cuqi/_version.py,sha256=qPKK2uUFBvtSRlgd2DlR8troFnOrSR3hAOTup08CdOY,510
 cuqi/config.py,sha256=wcYvz19wkeKW2EKCGIKJiTpWt5kdaxyt4imyRkvtTRA,526
 cuqi/diagnostics.py,sha256=5OrbJeqpynqRXOe5MtOKKhe7EAVdOEpHIqHnlMW9G_c,3029
 cuqi/array/__init__.py,sha256=-EeiaiWGNsE3twRS4dD814BIlfxEsNkTCZUc5gjOXb0,30
@@ -46,14 +46,14 @@ cuqi/experimental/mcmc/_langevin_algorithm.py,sha256=NIoCLKL5x89Bxm-JLDLR_NTunRE
 cuqi/experimental/mcmc/_laplace_approximation.py,sha256=XcGIa2wl9nCSTtAFurejYYOKkDVAJ22q75xQKsyu2nI,5803
 cuqi/experimental/mcmc/_mh.py,sha256=MXo0ahXP4KGFkaY4HtvcBE-TMQzsMlTmLKzSvpz7drU,2941
 cuqi/experimental/mcmc/_pcn.py,sha256=wqJBZLuRFSwxihaI53tumAg6AWVuceLMOmXssTetd1A,3374
-cuqi/experimental/mcmc/_rto.py,sha256=j3PD3ZfOuGifbBu51Z7GdMCM47HjH0luhpWFDPXNtxc,10477
+cuqi/experimental/mcmc/_rto.py,sha256=lzfeUuV8jUiWG-80KQ4if6toVcX7bMv-a0chBZq0vZ4,12021
 cuqi/experimental/mcmc/_sampler.py,sha256=BZHnpB6s-YSddd46wQSds0vNF61RA58Nc9ZU05WngdU,20184
 cuqi/experimental/mcmc/_utilities.py,sha256=kUzHbhIS3HYZRbneNBK41IogUYX5dS_bJxqEGm7TQBI,525
 cuqi/geometry/__init__.py,sha256=Tz1WGzZBY-QGH3c0GiyKm9XHN8MGGcnU6TUHLZkzB3o,842
-cuqi/geometry/_geometry.py,sha256=SDRZdiN2CIuS591lXxqgFoPWPIpwY-MHk75116QvdYY,46901
+cuqi/geometry/_geometry.py,sha256=5ZNrw6LivxEEw0vrk1eCxKIw-8mkAh7930voRVywDbY,47089
 cuqi/implicitprior/__init__.py,sha256=6z3lvw-tWDyjZSpB3pYzvijSMK9Zlf1IYqOVTtMD2h4,309
-cuqi/implicitprior/_regularizedGMRF.py,sha256=IR9tKzNMoz-b0RKu6ahVgMx_lDNB3jZHVWFMQm6QqZk,6259
-cuqi/implicitprior/_regularizedGaussian.py,sha256=cQtrgzyJU2pwoK4ORGl1erKLE9VY5NqwZTiqiViDswA,12371
+cuqi/implicitprior/_regularizedGMRF.py,sha256=rr3R2C1aheuu_KD35MureZKfOwY8O1pkVDHvuaFnFFU,6300
+cuqi/implicitprior/_regularizedGaussian.py,sha256=mzaAHq0yz73FZo-OB2iqFMd2i2NNzVv4mjd9-ger8a0,15435
 cuqi/implicitprior/_regularizedUnboundedUniform.py,sha256=H2fTOSqYTlDiLxQ7Ya6wnpCUIkpO4qKrkTOsOPnBBeU,3483
 cuqi/implicitprior/_restorator.py,sha256=ixnH8RGcLpqlaIUdR5Dwjx72sO9f3BeotNFRC7Z7qZo,9198
 cuqi/likelihood/__init__.py,sha256=QXif382iwZ5bT3ZUqmMs_n70JVbbjxbqMrlQYbMn4Zo,1776
@@ -81,14 +81,14 @@ cuqi/sampler/_sampler.py,sha256=TkZ_WAS-5Q43oICa-Elc2gftsRTBd7PEDUMDZ9tTGmU,5712
 cuqi/samples/__init__.py,sha256=vCs6lVk-pi8RBqa6cIN5wyn6u-K9oEf1Na4k1ZMrYv8,44
 cuqi/samples/_samples.py,sha256=hUc8OnCF9CTCuDTrGHwwzv3wp8mG_6vsJAFvuQ-x0uA,35832
 cuqi/solver/__init__.py,sha256=3eoTTgBHe3M6ygrbgUVG3GlqaZVe5lGajNV9rolXZJ8,179
-cuqi/solver/_solver.py,sha256=4LdfxLaU-fUHltZw7Sq-Xohyxd_6RvKy03xxtIMW6Zs,29488
+cuqi/solver/_solver.py,sha256=GquU_rj-9yfPQnBVE_gXo4wdF84xw_pLks3bJarzR58,29491
 cuqi/testproblem/__init__.py,sha256=DWTOcyuNHMbhEuuWlY5CkYkNDSAqhvsKmJXBLivyblU,202
 cuqi/testproblem/_testproblem.py,sha256=x769LwwRdJdzIiZkcQUGb_5-vynNTNALXWKato7sS0Q,52540
 cuqi/utilities/__init__.py,sha256=H7xpJe2UinjZftKvE2JuXtTi4DqtkR6uIezStAXwfGg,428
 cuqi/utilities/_get_python_variable_name.py,sha256=QwlBVj2koJRA8s8pWd554p7-ElcI7HUwY32HknaR92E,1827
 cuqi/utilities/_utilities.py,sha256=Jc4knn80vLoA7kgw9FzXwKVFGaNBOXiA9kgvltZU3Ao,11777
-CUQIpy-1.2.0.post0.dev305.dist-info/LICENSE,sha256=kJWRPrtRoQoZGXyyvu50Uc91X6_0XRaVfT0YZssicys,10799
-CUQIpy-1.2.0.post0.dev305.dist-info/METADATA,sha256=HMjktvjoT8Oo0N4v-2SYK5MnRGT6RdeqEwnQ2pMsAik,18496
-CUQIpy-1.2.0.post0.dev305.dist-info/WHEEL,sha256=R06PA3UVYHThwHvxuRWMqaGcr-PuniXahwjmQRFMEkY,91
-CUQIpy-1.2.0.post0.dev305.dist-info/top_level.txt,sha256=AgmgMc6TKfPPqbjV0kvAoCBN334i_Lwwojc7HE3ZwD0,5
-CUQIpy-1.2.0.post0.dev305.dist-info/RECORD,,
+CUQIpy-1.2.0.post0.dev333.dist-info/LICENSE,sha256=kJWRPrtRoQoZGXyyvu50Uc91X6_0XRaVfT0YZssicys,10799
+CUQIpy-1.2.0.post0.dev333.dist-info/METADATA,sha256=01FrLb_xnRL-lDXDLU48mtTy--nELMrPyY_rX0deLUg,18496
+CUQIpy-1.2.0.post0.dev333.dist-info/WHEEL,sha256=R06PA3UVYHThwHvxuRWMqaGcr-PuniXahwjmQRFMEkY,91
+CUQIpy-1.2.0.post0.dev333.dist-info/top_level.txt,sha256=AgmgMc6TKfPPqbjV0kvAoCBN334i_Lwwojc7HE3ZwD0,5
+CUQIpy-1.2.0.post0.dev333.dist-info/RECORD,,

cuqi/_version.py

@@ -8,11 +8,11 @@ import json
 
 version_json = '''
 {
- "date": "2024-11-19T23:04:41+0300",
+ "date": "2024-11-20T12:41:18+0100",
  "dirty": false,
  "error": null,
- "full-revisionid": "40cfaf6a083e595a757ef044e21ab931a869b2fe",
- "version": "1.2.0.post0.dev305"
+ "full-revisionid": "e495208f39ca278465cb4819799b22c942a1b863",
+ "version": "1.2.0.post0.dev333"
 }
 '''  # END VERSION_JSON
 

cuqi/experimental/mcmc/_rto.py

@@ -3,7 +3,7 @@ from scipy.linalg.interpolative import estimate_spectral_norm
 from scipy.sparse.linalg import LinearOperator as scipyLinearOperator
 import numpy as np
 import cuqi
-from cuqi.solver import CGLS, FISTA
+from cuqi.solver import CGLS, FISTA, ADMM
 from cuqi.experimental.mcmc import Sampler
 
 
@@ -161,6 +161,13 @@ class RegularizedLinearRTO(LinearRTO):
     Regularized Linear RTO (Randomize-Then-Optimize) sampler.
 
     Samples posterior related to the inverse problem with Gaussian likelihood and implicit Gaussian prior, and where the forward model is Linear.
+    The sampler works by repeatedly solving regularized linear least squares problems for perturbed data.
+    The solver for these optimization problems is chosen based on how the regularized is provided in the implicit Gaussian prior.
+    Currently we use the following solvers:
+    FISTA: [1] Beck, Amir, and Marc Teboulle. "A fast iterative shrinkage-thresholding algorithm for linear inverse problems." SIAM journal on imaging sciences 2.1 (2009): 183-202.
+           Used when prior.proximal is callable.
+    ADMM:  [2] Boyd et al. "Distributed optimization and statistical learning via the alternating direction method of multipliers."Foundations and Trends® in Machine learning, 2011.
+           Used when prior.proximal is a list of penalty terms.
 
     Parameters
     ------------
@@ -171,12 +178,19 @@ class RegularizedLinearRTO(LinearRTO):
         Initial point for the sampler. *Optional*.
 
     maxit : int
-        Maximum number of iterations of the inner FISTA solver. *Optional*.
+        Maximum number of iterations of the inner FISTA/ADMM solver. *Optional*.
+
+    inner_max_it : int
+        Maximum number of iterations of the CGLS solver used within the ADMM solver. *Optional*.
         
     stepsize : string or float
         If stepsize is a string and equals either "automatic", then the stepsize is automatically estimated based on the spectral norm.
         If stepsize is a float, then this stepsize is used.
 
+    penalty_parameter : int
+        Penalty parameter of the inner ADMM solver. *Optional*.
+        See [2] or `cuqi.solver.ADMM`
+
     abstol : float
         Absolute tolerance of the inner FISTA solver. *Optional*.
     
@@ -190,7 +204,7 @@ class RegularizedLinearRTO(LinearRTO):
         An example is shown in demos/demo31_callback.py.
         
     """
-    def __init__(self, target=None, initial_point=None, maxit=100, stepsize="automatic", abstol=1e-10, adaptive=True, **kwargs):
+    def __init__(self, target=None, initial_point=None, maxit=100, inner_max_it=10, stepsize="automatic", penalty_parameter=10, abstol=1e-10, adaptive=True, **kwargs):
         
         super().__init__(target=target, initial_point=initial_point, **kwargs)
 
@@ -199,10 +213,13 @@ class RegularizedLinearRTO(LinearRTO):
         self.abstol = abstol   
         self.adaptive = adaptive
         self.maxit = maxit
+        self.inner_max_it = inner_max_it
+        self.penalty_parameter = penalty_parameter
 
     def _initialize(self):
         super()._initialize()
-        self._stepsize = self._choose_stepsize()
+        if self._inner_solver == "FISTA":
+            self._stepsize = self._choose_stepsize()
 
     @property
     def proximal(self):
@@ -212,8 +229,7 @@ class RegularizedLinearRTO(LinearRTO):
         super().validate_target()
         if not isinstance(self.target.prior, (cuqi.implicitprior.RegularizedGaussian, cuqi.implicitprior.RegularizedGMRF)):
             raise TypeError("Prior needs to be RegularizedGaussian or RegularizedGMRF")
-        if not callable(self.proximal):
-            raise TypeError("Proximal needs to be callable")
+        self._inner_solver = "FISTA" if callable(self.proximal) else "ADMM"
 
     def _choose_stepsize(self):
         if isinstance(self.stepsize, str):
@@ -237,8 +253,16 @@ class RegularizedLinearRTO(LinearRTO):
 
     def step(self):
         y = self.b_tild + np.random.randn(len(self.b_tild))
-        sim = FISTA(self.M, y, self.proximal,
-                    self.current_point, maxit = self.maxit, stepsize = self._stepsize, abstol = self.abstol, adaptive = self.adaptive)         
+
+        if self._inner_solver == "FISTA":
+            sim = FISTA(self.M, y, self.proximal,
+                        self.current_point, maxit = self.maxit, stepsize = self._stepsize, abstol = self.abstol, adaptive = self.adaptive)         
+        elif self._inner_solver == "ADMM":
+            sim = ADMM(self.M, y, self.proximal,
+                        self.current_point, self.penalty_parameter, maxit = self.maxit, inner_max_it = self.inner_max_it, adaptive = self.adaptive)  
+        else:
+            raise ValueError("Choice of solver not supported.")
+
         self.current_point, _ = sim.solve()
         acc = 1
         return acc

cuqi/geometry/_geometry.py

@@ -522,6 +522,9 @@ class Image2D(Geometry):
     Plotting is handled via matplotlib.pyplot.imshow.
     Colormap is defaulted to grayscale.
 
+    A How-To guide on the use of Image2D as the domain/range geometry of a CUQI 
+    :class:`Model` is available `here <https://cuqi-dtu.github.io/CUQIpy/user/_auto_howtos/myula.html>`_.
+
     Parameters
     -----------
     im_shape : tuple

cuqi/implicitprior/_regularizedGMRF.py

@@ -63,6 +63,7 @@ class RegularizedGMRF(RegularizedGaussian):
             
             # Underlying explicit GMRF
             self._gaussian = GMRF(mean, prec, bc_type=bc_type, order=order, **kwargs)
+            kwargs.pop("geometry", None)
             
             # Init from abstract distribution class
             super(Distribution, self).__init__(**kwargs)

cuqi/implicitprior/_regularizedGaussian.py

@@ -1,6 +1,8 @@
 from cuqi.utilities import get_non_default_args
 from cuqi.distribution import Distribution, Gaussian
 from cuqi.solver import ProjectNonnegative, ProjectBox, ProximalL1
+from cuqi.geometry import Continuous1D, Continuous2D, Image2D
+from cuqi.operator import FirstOrderFiniteDifference
 
 import numpy as np
 
@@ -39,17 +41,22 @@ class RegularizedGaussian(Distribution):
     sqrtprec
         See :class:`~cuqi.distribution.Gaussian` for details.
 
-    proximal : callable f(x, scale) or None
-        Euclidean proximal operator f of the regularization function g, that is, a solver for the optimization problem
-        min_z 0.5||x-z||_2^2+scale*g(x).
-
+    proximal : callable f(x, scale), list of tuples (callable proximal operator of f_i, linear operator L_i) or None
+        If callable:
+            Euclidean proximal operator f of the regularization function g, that is, a solver for the optimization problem
+            min_z 0.5||x-z||_2^2+scale*g(x).
+        If list of tuples (callable proximal operator of f_i, linear operator L_i):
+            Each callable proximal operator of f_i accepts two arguments (x, p) and should return the minimizer of p/2||x-z||^2 + f(x) over z for some f.
+            The corresponding regularization takes the form
+                sum_i f_i(L_i x),
+            where the sum ranges from 1 to an arbitrary n.
 
     projector : callable f(x) or None
         Euclidean projection onto the constraint C, that is, a solver for the optimization problem
         min_(z in C) 0.5||x-z||_2^2.
 
     constraint : string or None
-        Preset constraints. Can be set to "nonnegativity" and "box". Required for use in Gibbs.
+        Preset constraints that generate the corresponding proximal parameter. Can be set to "nonnegativity" and "box". Required for use in Gibbs.
         For "box", the following additional parameters can be passed:
             lower_bound : array_like or None
                 Lower bound of box, defaults to zero
@@ -57,10 +64,10 @@ class RegularizedGaussian(Distribution):
                 Upper bound of box, defaults to one
 
     regularization : string or None
-        Preset regularization. Can be set to "l1". Required for use in Gibbs in future update.
-        For "l1", the following additional parameters can be passed:
+        Preset regularization that generate the corresponding proximal parameter. Can be set to "l1" or 'tv'. Required for use in Gibbs in future update.
+        For "l1" or "tv", the following additional parameters can be passed:
             strength : scalar
-                Regularization parameter, i.e., strength*||x||_1 , defaults to one
+                Regularization parameter, i.e., strength*||Lx||_1, defaults to one
 
     """
         
@@ -75,6 +82,7 @@ class RegularizedGaussian(Distribution):
         
         # We init the underlying Gaussian first for geometry and dimensionality handling
         self._gaussian = Gaussian(mean=mean, cov=cov, prec=prec, sqrtcov=sqrtcov, sqrtprec=sqrtprec, **kwargs)
+        kwargs.pop("geometry", None)
 
         # Init from abstract distribution class
         super().__init__(**kwargs)
@@ -88,12 +96,6 @@ class RegularizedGaussian(Distribution):
         if (proximal is not None) + (projector is not None) + (constraint is not None) + (regularization is not None) != 1:
             raise ValueError("Precisely one of proximal, projector, constraint or regularization needs to be provided.")
 
-        if proximal is not None:
-            if not callable(proximal):
-                raise ValueError("Proximal needs to be callable.")
-            if len(get_non_default_args(proximal)) != 2:
-                raise ValueError("Proximal should take 2 arguments.")
-            
         if projector is not None:
             if not callable(projector):
                 raise ValueError("Projector needs to be callable.")
@@ -104,7 +106,8 @@ class RegularizedGaussian(Distribution):
         self._preset = None
         
         if proximal is not None:
-            self._proximal = proximal
+            # No need to generate the proximal and associated information
+            self.proximal = proximal
         elif projector is not None:
             self._proximal = lambda z, gamma: projector(z)
         elif (isinstance(constraint, str) and constraint.lower() == "nonnegativity"):
@@ -113,15 +116,48 @@ class RegularizedGaussian(Distribution):
         elif (isinstance(constraint, str) and constraint.lower() == "box"):
             lower = optional_regularization_parameters["lower_bound"]
             upper = optional_regularization_parameters["upper_bound"]
-            self._proximal = lambda z, gamma: ProjectBox(z, lower, upper)
+            self._proximal = lambda z, _: ProjectBox(z, lower, upper)
             self._preset = "box" # Not supported in Gibbs
         elif (isinstance(regularization, str) and regularization.lower() in ["l1"]):
-            strength = optional_regularization_parameters["strength"]
-            self._proximal = lambda z, gamma: ProximalL1(z, gamma*strength)
+            self._strength = optional_regularization_parameters["strength"]
+            self._proximal = lambda z, gamma: ProximalL1(z, gamma*self._strength)
             self._preset = "l1"
+        elif (isinstance(regularization, str) and regularization.lower() in ["tv"]):
+            self._strength = optional_regularization_parameters["strength"]
+            if isinstance(self.geometry, (Continuous1D, Continuous2D, Image2D)):
+                self._transformation = FirstOrderFiniteDifference(self.geometry.fun_shape, bc_type='neumann')
+            else:
+                raise ValueError("Geometry not supported for total variation")
+            
+            self._regularization_prox = lambda z, gamma: ProximalL1(z, gamma*self._strength)
+            self._regularization_oper = self._transformation
+
+            self._proximal = [(self._regularization_prox, self._regularization_oper)]
+            self._preset = "tv"
         else:
             raise ValueError("Regularization not supported")
 
+    
+    @property
+    def transformation(self):
+        return self._transformation
+    
+    @property
+    def strength(self):
+        return self._strength
+        
+    @strength.setter
+    def strength(self, value):
+        if self._preset not in self.regularization_options():
+            raise TypeError("Strength is only used when the regularization is set to l1 or TV.")
+
+        self._strength = value
+        if self._preset == "tv":
+            self._regularization_prox = lambda z, gamma: ProximalL1(z, gamma*self._strength)
+            self._proximal = [(self._regularization_prox, self._regularization_oper)]
+        elif self._preset == "l1":
+            self._proximal = lambda z, gamma: ProximalL1(z, gamma*self._strength)
+
     # This is a getter only attribute for the underlying Gaussian
     # It also ensures that the name of the underlying Gaussian
     # matches the name of the implicit regularized Gaussian
@@ -135,6 +171,25 @@ class RegularizedGaussian(Distribution):
     def proximal(self):
         return self._proximal
     
+    @proximal.setter
+    def proximal(self, value):
+        if callable(value):
+            if len(get_non_default_args(value)) != 2:
+                raise ValueError("Proximal should take 2 arguments.")
+        elif isinstance(value, list):
+            for (prox, op) in value:
+                if len(get_non_default_args(prox)) != 2:
+                    raise ValueError("Proximal should take 2 arguments.")
+                if op.shape[1] != self.geometry.par_dim:
+                    raise ValueError("Incorrect shape of linear operator in proximal list.")
+        else:
+            raise ValueError("Proximal needs to be callable or a list. See documentation.")
+        
+        self._proximal = value
+
+        # For all the presets, self._proximal is set directly, 
+        self._preset = None
+            
     @property
     def preset(self):
         return self._preset
@@ -154,7 +209,7 @@ class RegularizedGaussian(Distribution):
 
     @staticmethod
     def regularization_options():
-        return ["l1"]
+        return ["l1", "tv"]
 
 
     # --- Defer behavior of the underlying Gaussian --- #
@@ -206,16 +261,18 @@ class RegularizedGaussian(Distribution):
     def sqrtcov(self, value):
         self.gaussian.sqrtcov = value     
     
-    def get_conditioning_variables(self):
-        return self.gaussian.get_conditioning_variables()
-    
     def get_mutable_variables(self):
-        return self.gaussian.get_mutable_variables()
+        mutable_vars = self.gaussian.get_mutable_variables().copy()
+        if self.preset in self.regularization_options():
+            mutable_vars += ["strength"]
+        return mutable_vars
     
     # Overwrite the condition method such that the underlying Gaussian is conditioned in general, except when conditioning on self.name
     # which means we convert Distribution to Likelihood or EvaluatedDensity.
     def _condition(self, *args, **kwargs):
-
+        if self.preset in self.regularization_options():
+            return super()._condition(*args, **kwargs)
+        
         # Handle positional arguments (similar code as in Distribution._condition)
         cond_vars = self.get_conditioning_variables()
         kwargs = self._parse_args_add_to_kwargs(cond_vars, *args, **kwargs)
@@ -275,7 +332,7 @@ class ConstrainedGaussian(RegularizedGaussian):
         min_(z in C) 0.5||x-z||_2^2.
 
     constraint : string or None
-        Preset constraints. Can be set to "nonnegativity" and "box". Required for use in Gibbs.
+        Preset constraints that generate the corresponding proximal parameter. Can be set to "nonnegativity" and "box". Required for use in Gibbs.
         For "box", the following additional parameters can be passed:
             lower_bound : array_like or None
                 Lower bound of box, defaults to zero

cuqi/solver/_solver.py

@@ -669,7 +669,7 @@ class ADMM(object):
         - flag=2 indicates multiplication of the transpose of A with vector x, that is  A.T @ x.
     b : ndarray.
     penalty_terms : List of tuples (callable proximal operator of f_i, linear operator L_i)
-        Each callable proximal operator f_i accepts two arguments (x, p) and should return the minimizer of p/2||x-z||^2 + f(x) over z for some f.
+        Each callable proximal operator of f_i accepts two arguments (x, p) and should return the minimizer of p/2||x-z||^2 + f(x) over z for some f.
     x0 : ndarray. Initial guess.
     penalty_parameter : Trade-off between linear least squares and regularization term in the solver iterates. Denoted as "rho" in [1].
     maxit : The maximum number of iterations.