CUQIpy 1.0.0.post0.dev215__tar.gz → 1.0.0.post0.dev305__tar.gz

{cuqipy-1.0.0.post0.dev215 → cuqipy-1.0.0.post0.dev305}/CUQIpy.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: CUQIpy
-Version: 1.0.0.post0.dev215
+Version: 1.0.0.post0.dev305
 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-1.0.0.post0.dev215 → cuqipy-1.0.0.post0.dev305}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: CUQIpy
-Version: 1.0.0.post0.dev215
+Version: 1.0.0.post0.dev305
 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-1.0.0.post0.dev215 → cuqipy-1.0.0.post0.dev305}/cuqi/_version.py

@@ -8,11 +8,11 @@ import json
 
 version_json = '''
 {
- "date": "2024-05-16T11:29:50+0200",
+ "date": "2024-05-27T07:34:21+0200",
  "dirty": false,
  "error": null,
- "full-revisionid": "b713d6478d33977b12342e4bcb29968a95ee5973",
- "version": "1.0.0.post0.dev215"
+ "full-revisionid": "1274ccfff1cb9663b547dd2aeb9c86c72dccb2ae",
+ "version": "1.0.0.post0.dev305"
 }
 '''  # END VERSION_JSON
 

{cuqipy-1.0.0.post0.dev215 → cuqipy-1.0.0.post0.dev305}/cuqi/experimental/mcmc/__init__.py

@@ -3,7 +3,7 @@
 from ._sampler import SamplerNew, ProposalBasedSamplerNew
 from ._langevin_algorithm import ULANew, MALANew
 from ._mh import MHNew
-from ._pcn import pCNNew
+from ._pcn import PCNNew
 from ._rto import LinearRTONew, RegularizedLinearRTONew
 from ._cwmh import CWMHNew
 from ._laplace_approximation import UGLANew

{cuqipy-1.0.0.post0.dev215 → cuqipy-1.0.0.post0.dev305}/cuqi/experimental/mcmc/_cwmh.py

@@ -67,16 +67,21 @@ class CWMHNew(ProposalBasedSamplerNew):
         samples = sampler.sample(2000).get_samples()
 
     """
-    def __init__(self, target: cuqi.density.Density, proposal=None, scale=1,
+
+    _STATE_KEYS = ProposalBasedSamplerNew._STATE_KEYS.union(['_scale_temp'])
+
+    def __init__(self, target:cuqi.density.Density=None, proposal=None, scale=1,
                  initial_point=None, **kwargs):
         super().__init__(target, proposal=proposal, scale=scale,
                          initial_point=initial_point, **kwargs)
+        
+    def _initialize(self):
+        if isinstance(self.scale, Number):
+            self.scale = np.ones(self.dim)*self.scale
+        self._acc = [np.ones((self.dim))] # Overwrite acc from ProposalBasedSamplerNew with list of arrays
 
-        # set initial scale
-        self.scale = scale
-
-        # set initial acceptance rate
-        self._acc = [np.ones((self.dim))]
+        # Handling of temporary scale parameter due to possible bug in old CWMH
+        self._scale_temp = self.scale.copy()
 
     @property
     def scale(self):
@@ -86,53 +91,35 @@ class CWMHNew(ProposalBasedSamplerNew):
     @scale.setter
     def scale(self, value):
         """ Set the scale parameter. """
-        if isinstance(value, Number):
-            self._scale = np.ones(self.dim)*value
-        elif isinstance(value, np.ndarray):
-            self._scale = value
-        self._scale_temp = self._scale.copy()
+        if self._is_initialized and isinstance(value, Number):
+            value = np.ones(self.dim)*value
+        self._scale = value
 
     def validate_target(self):
         if not isinstance(self.target, cuqi.density.Density):
             raise ValueError(
                 "Target should be an instance of "+\
                 f"{cuqi.density.Density.__class__.__name__}")
-
-    @ProposalBasedSamplerNew.proposal.setter
-    # TODO. Check if we can refactor this.
-    # We can work with a validate_proposal method instead?
-    def proposal(self, value):
-        fail_msg = "Proposal should be either None, "+\
-            f"{cuqi.distribution.Distribution.__class__.__name__} "+\
-            "conditioned only on 'location' and 'scale', lambda function, "+\
-            f"or {cuqi.distribution.Normal.__class__.__name__} conditioned "+\
-            "only on 'mean' and 'std'"
-
-        if value is None:
+        
+    def validate_proposal(self):
+        if not isinstance(self.proposal, cuqi.distribution.Distribution):
+            raise ValueError("Proposal must be a cuqi.distribution.Distribution object")
+        if not self.proposal.is_symmetric:
+            raise ValueError("Proposal must be symmetric")
+        
+    @property
+    def proposal(self):
+        if self._proposal is None:
             self._proposal = cuqi.distribution.Normal(
                 mean=lambda location: location,
                 std=lambda scale: scale,
                 geometry=self.dim,
             )
-
-        elif isinstance(value, cuqi.distribution.Distribution) and sorted(
-            value.get_conditioning_variables()
-        ) == ["location", "scale"]:
-            self._proposal = value
-
-        elif isinstance(value, cuqi.distribution.Normal) and sorted(
-            value.get_conditioning_variables()
-        ) == ["mean", "std"]:
-            self._proposal = value(
-                mean=lambda location: location, std=lambda scale: scale
-            )
-
-        elif not isinstance(value, cuqi.distribution.Distribution) and callable(
-            value):
-            self._proposal = value
-
-        else:
-            raise ValueError(fail_msg)
+        return self._proposal
+    
+    @proposal.setter
+    def proposal(self, value):
+        self._proposal = value
 
     def step(self):
         # Initialize x_t which is used to store the current CWMH sample

{cuqipy-1.0.0.post0.dev215 → cuqipy-1.0.0.post0.dev305}/cuqi/experimental/mcmc/_hmc.py

@@ -95,7 +95,7 @@ class NUTSNew(SamplerNew):
                                                     'epsilon_list',
                                                     'epsilon_bar_list'})
 
-    def __init__(self, target, initial_point=None, max_depth=15,
+    def __init__(self, target=None, initial_point=None, max_depth=15,
                  step_size=None, opt_acc_rate=0.6, **kwargs):
         super().__init__(target, initial_point=initial_point, **kwargs)
 
@@ -103,9 +103,6 @@ class NUTSNew(SamplerNew):
         self.max_depth = max_depth
         self.step_size = step_size
         self.opt_acc_rate = opt_acc_rate
-        
-        # Set current point 
-        self.current_point = self.initial_point
 
         # Initialize epsilon and epsilon_bar
         # epsilon is the step size used in the current iteration
@@ -115,12 +112,37 @@ class NUTSNew(SamplerNew):
         self._epsilon_bar = None
         self._H_bar = None
 
+        # Extra parameters for tuning
+        self._n_alpha = None
+        self._alpha = None
+
+
+    def _initialize(self):
+
         # Arrays to store acceptance rate
-        self._acc = [None]
+        self._acc = [None] # Overwrites acc from SamplerNew. TODO. Check if this is necessary
+
+        self._alpha = 0 # check if meaningful value
+        self._n_alpha = 0 # check if meaningful value
+
+        self.current_target_logd, self.current_target_grad = self._nuts_target(self.current_point)
+
+        # parameters dual averaging
+        if self.step_size is None:
+            self._epsilon = self._FindGoodEpsilon()
+        else:
+            self._epsilon = self.step_size
+        self._epsilon_bar = "unset"
+
+        # Parameter mu, does not change during the run
+        self._mu = np.log(10*self._epsilon)
+
+        self._H_bar = 0
 
         # NUTS run diagnostic:
         # number of tree nodes created each NUTS iteration
         self._num_tree_node = 0
+
         # Create lists to store NUTS run diagnostics
         self._create_run_diagnostic_attributes()
 
@@ -176,9 +198,9 @@ class NUTSNew(SamplerNew):
         except:
             raise ValueError('Target must have logd and gradient methods.')
 
-    def reset(self):
+    def reinitialize(self):
         # Call the parent reset method
-        super().reset()
+        super().reinitialize()
         # Reset NUTS run diagnostic attributes
         self._reset_run_diagnostic_attributes()
 
@@ -275,10 +297,6 @@ class NUTSNew(SamplerNew):
             np.exp(eta*np.log(self._epsilon) +(1-eta)*np.log(self._epsilon_bar))
 
     def _pre_warmup(self):
-        super()._pre_warmup()
-
-        self.current_target_logd, self.current_target_grad =\
-            self._nuts_target(self.current_point)
 
         # Set up tuning parameters (only first time tuning is called)
         # Note:
@@ -289,32 +307,14 @@ class NUTSNew(SamplerNew):
         #  Parameters that does not change during the run
         #    self._mu
 
-        if self._epsilon is None:
-            # parameters dual averaging
-            self._epsilon = self._FindGoodEpsilon()
-            # Parameter mu, does not change during the run
-            self._mu = np.log(10*self._epsilon)
-
-        if self._epsilon_bar is None: # Initial value of epsilon_bar
+        if self._epsilon_bar == "unset": # Initial value of epsilon_bar for tuning
             self._epsilon_bar = 1
 
-        if self._H_bar is None: # Initial value of H_bar
-            self._H_bar = 0
-
     def _pre_sample(self):
-        super()._pre_sample()
 
-        self.current_target_logd, self.current_target_grad =\
-            self._nuts_target(self.current_point)
-        
-        # Set up epsilon and epsilon_bar if not set
-        if self._epsilon is None:
-            if self.step_size is None:
-                step_size = self._FindGoodEpsilon()
-            else:
-                step_size = self.step_size
-            self._epsilon = step_size
-            self._epsilon_bar = step_size
+        if self._epsilon_bar == "unset": # Initial value of epsilon_bar for sampling
+            self._epsilon_bar = self._epsilon
+            
 
     #=========================================================================
     def _nuts_target(self, x): # returns logposterior tuple evaluation-gradient
@@ -467,4 +467,4 @@ class NUTSNew(SamplerNew):
         # Store the step size used in iteration k
         self.epsilon_list.append(eps)
         # Store the step size suggestion during adaptation in iteration k
-        self.epsilon_bar_list.append(eps_bar)
+        self.epsilon_bar_list.append(eps_bar)

{cuqipy-1.0.0.post0.dev215 → cuqipy-1.0.0.post0.dev305}/cuqi/experimental/mcmc/_langevin_algorithm.py

@@ -63,15 +63,16 @@ class ULANew(SamplerNew): # Refactor to Proposal-based sampler?
 
     _STATE_KEYS = SamplerNew._STATE_KEYS.union({'current_target_logd', 'scale', 'current_target_grad'})
 
-    def __init__(self, target, scale=1.0, **kwargs):
+    def __init__(self, target=None, scale=1.0, **kwargs):
 
         super().__init__(target, **kwargs)
 
-        self.scale = scale
-        self.current_point = self.initial_point
+        self.initial_scale = scale
+
+    def _initialize(self):
+        self.scale = self.initial_scale
         self.current_target_logd = self.target.logd(self.current_point)
         self.current_target_grad = self.target.gradient(self.current_point)
-        self._acc = [1] # TODO. Check if we need this
 
     def validate_target(self):
         try:

{cuqipy-1.0.0.post0.dev215 → cuqipy-1.0.0.post0.dev305}/cuqi/experimental/mcmc/_laplace_approximation.py

@@ -53,17 +53,13 @@ class UGLANew(SamplerNew):
 
         super().__init__(target=target, initial_point=initial_point, **kwargs)
 
-        if initial_point is None: #TODO: Replace later with a getter
-            self.initial_point = np.zeros(self.dim)
-            self._samples = [self.initial_point]
-
-        self.current_point = self.initial_point
-        self._acc = [1] # TODO. Check if we need this
-
         # Parameters
         self.maxit = maxit
         self.tol = tol
         self.beta = beta
+    
+    def _initialize(self):
+        self._precompute()
 
     @property
     def prior(self):
@@ -81,8 +77,8 @@ class UGLANew(SamplerNew):
     def data(self):
         return self.target.data
 
-    def _pre_warmup(self):
-        super()._pre_warmup()
+    def _precompute(self):
+
         D = self.prior._diff_op
         n = D.shape[0]
 
@@ -121,9 +117,6 @@ class UGLANew(SamplerNew):
             return out
         self.M = M
 
-    def _pre_sample(self):
-        self._pre_warmup()
-
     def step(self):
         # Update Laplace approximation
         self._L2 = self.Lk_fun(self.current_point)
@@ -157,3 +150,8 @@ class UGLANew(SamplerNew):
         # Check that prior is LMRF
         if not isinstance(self.prior, cuqi.distribution.LMRF):
             raise ValueError('Unadjusted Gaussian Laplace approximation (UGLA) requires LMRF prior')
+        
+    @property
+    def _default_initial_point(self):
+        """ Get the default initial point for the sampler. Defaults to an array of zeros. """
+        return np.zeros(self.dim)

{cuqipy-1.0.0.post0.dev215 → cuqipy-1.0.0.post0.dev305}/cuqi/experimental/mcmc/_mh.py

@@ -1,7 +1,6 @@
 import numpy as np
 import cuqi
 from cuqi.experimental.mcmc import ProposalBasedSamplerNew
-from cuqi.array import CUQIarray
 
 
 class MHNew(ProposalBasedSamplerNew):
@@ -27,25 +26,19 @@ class MHNew(ProposalBasedSamplerNew):
 
     def __init__(self, target, proposal=None, scale=1, **kwargs):
         super().__init__(target, proposal=proposal, scale=scale, **kwargs)
+
+    def _initialize(self):
         # Due to a bug? in old MH, we must keep track of this extra variable to match behavior.
-        self._scale_temp = self.scale 
+        self._scale_temp = self.scale
 
     def validate_target(self):
         pass # All targets are valid
 
-    @ProposalBasedSamplerNew.proposal.setter  # TODO. Check if we can refactor this. We can work with a validate_proposal method instead?
-    def proposal(self, value):
-        fail_msg = "Proposal should be either None, symmetric cuqi.distribution.Distribution or a lambda function."
-
-        if value is None:
-            self._proposal = cuqi.distribution.Gaussian(np.zeros(self.dim), 1)
-        elif not isinstance(value, cuqi.distribution.Distribution) and callable(value):
-            raise NotImplementedError(fail_msg)
-        elif isinstance(value, cuqi.distribution.Distribution) and value.is_symmetric:
-            self._proposal = value
-        else:
-            raise ValueError(fail_msg)
-        self._proposal.geometry = self.target.geometry
+    def validate_proposal(self):
+        if not isinstance(self.proposal, cuqi.distribution.Distribution):
+            raise ValueError("Proposal must be a cuqi.distribution.Distribution object")
+        if not self.proposal.is_symmetric:
+            raise ValueError("Proposal must be symmetric")
 
     def step(self):
         # propose state

cuqipy-1.0.0.post0.dev305/cuqi/experimental/mcmc/_pcn.py

@@ -0,0 +1,89 @@
+import numpy as np
+import cuqi
+from cuqi.experimental.mcmc import SamplerNew
+from cuqi.array import CUQIarray
+
+class PCNNew(SamplerNew):  # Refactor to Proposal-based sampler?
+
+    _STATE_KEYS = SamplerNew._STATE_KEYS.union({'scale', 'current_likelihood_logd', 'lambd'})
+
+    def __init__(self, target=None, scale=1.0, **kwargs):
+
+        super().__init__(target, **kwargs)
+        self.initial_scale = scale
+
+    def _initialize(self):
+        self.scale = self.initial_scale
+        self.current_likelihood_logd = self._loglikelihood(self.current_point)
+
+        # parameters used in the Robbins-Monro recursion for tuning the scale parameter
+        # see details and reference in the tune method
+        self.lambd = self.scale
+        self.star_acc = 0.44 #TODO: 0.234 # target acceptance rate
+
+    def validate_target(self):
+        if not isinstance(self.target, cuqi.distribution.Posterior):
+            raise ValueError(f"To initialize an object of type {self.__class__}, 'target' need to be of type 'cuqi.distribution.Posterior'.")
+        if not isinstance(self.prior, (cuqi.distribution.Gaussian, cuqi.distribution.Normal)):
+            raise ValueError("The prior distribution of the target need to be Gaussian")
+
+    def step(self):
+        # propose state
+        xi = self.prior.sample(1).flatten()   # sample from the prior
+        x_star = np.sqrt(1-self.scale**2)*self.current_point + self.scale*xi   # PCN proposal
+
+        # evaluate target
+        loglike_eval_star =  self._loglikelihood(x_star) 
+
+        # ratio and acceptance probability
+        ratio = loglike_eval_star - self.current_likelihood_logd  # proposal is symmetric
+        alpha = min(0, ratio)
+
+        # accept/reject
+        acc = 0
+        u_theta = np.log(np.random.rand())
+        if (u_theta <= alpha):
+            self.current_point = x_star
+            self.current_likelihood_logd = loglike_eval_star
+            acc = 1
+        
+        return acc
+
+    @property
+    def prior(self):
+        return self.target.prior
+
+    @property
+    def likelihood(self):
+        return self.target.likelihood
+        
+    def _loglikelihood(self, x):
+        return self.likelihood.logd(x)
+
+    @property
+    def dim(self): # TODO. Check if we need this. Implemented in base class
+        if hasattr(self,'target') and hasattr(self.target,'dim'):
+            self._dim = self.target.dim
+        elif hasattr(self,'target') and isinstance(self.target,tuple) and len(self.target)==2:
+            self._dim = self.target[0].dim
+        return self._dim
+
+    def tune(self, skip_len, update_count):
+        """
+        Tune the scale parameter of the PCN sampler.
+        The tuning is based on algorithm 4 in Andrieu, Christophe, and Johannes Thoms. 
+        "A tutorial on adaptive MCMC." Statistics and computing 18 (2008): 343-373.
+        Note: the tuning algorithm here is the same as the one used in MH sampler.
+        """
+
+        # average acceptance rate in the past skip_len iterations
+        hat_acc = np.mean(self._acc[-skip_len:])
+
+        # new scaling parameter zeta to be used in the Robbins-Monro recursion
+        zeta = 1/np.sqrt(update_count+1)
+
+        # Robbins-Monro recursion to ensure that the variation of lambd vanishes
+        self.lambd = np.exp(np.log(self.lambd) + zeta*(hat_acc-self.star_acc))
+
+        # update scale parameter
+        self.scale = min(self.lambd, 1)

{cuqipy-1.0.0.post0.dev215 → cuqipy-1.0.0.post0.dev305}/cuqi/experimental/mcmc/_rto.py

@@ -5,7 +5,6 @@ import numpy as np
 import cuqi
 from cuqi.solver import CGLS, FISTA
 from cuqi.experimental.mcmc import SamplerNew
-from cuqi.array import CUQIarray
 
 
 class LinearRTONew(SamplerNew):
@@ -44,21 +43,17 @@ class LinearRTONew(SamplerNew):
         An example is shown in demos/demo31_callback.py.
         
     """
-    def __init__(self, target, initial_point=None, maxit=10, tol=1e-6, **kwargs):
+    def __init__(self, target=None, initial_point=None, maxit=10, tol=1e-6, **kwargs):
 
         super().__init__(target=target, initial_point=initial_point, **kwargs)
 
-        if initial_point is None: #TODO: Replace later with a getter
-            self.initial_point = np.zeros(self.dim)
-            self._samples = [self.initial_point]
-
-        self.current_point = self.initial_point
-        self._acc = [1] # TODO. Check if we need this
-
         # Other parameters
         self.maxit = maxit
         self.tol = tol
 
+    def _initialize(self):
+        self._precompute()
+
     @property
     def prior(self):
         return self.target.prior
@@ -81,41 +76,7 @@ class LinearRTONew(SamplerNew):
     @property
     def data(self):
         return self.target.data
-
-    @SamplerNew.target.setter
-    def target(self, value):
-        """ Set the target density. Runs validation of the target. """
-        # Accept tuple of inputs and construct posterior
-        if isinstance(value, tuple) and len(value) == 5:
-            # Structure (data, model, L_sqrtprec, P_mean, P_sqrtprec)
-            data = value[0]
-            model = value[1]
-            L_sqrtprec = value[2]
-            P_mean = value[3]
-            P_sqrtprec = value[4]
-
-            # If numpy matrix convert to CUQI model
-            if isinstance(model, np.ndarray) and len(model.shape) == 2:
-                model = cuqi.model.LinearModel(model)
-
-            # Check model input
-            if not isinstance(model, cuqi.model.LinearModel):
-                raise TypeError("Model needs to be cuqi.model.LinearModel or matrix")
-
-            # Likelihood
-            L = cuqi.distribution.Gaussian(model, sqrtprec=L_sqrtprec).to_likelihood(data)
-
-            # Prior TODO: allow multiple priors stacked
-            #if isinstance(P_mean, list) and isinstance(P_sqrtprec, list):
-            #    P = cuqi.distribution.JointGaussianSqrtPrec(P_mean, P_sqrtprec)
-            #else:
-            P = cuqi.distribution.Gaussian(P_mean, sqrtprec=P_sqrtprec)
-
-            # Construct posterior
-            value = cuqi.distribution.Posterior(L, P)
-        super(LinearRTONew, type(self)).target.fset(self, value)
-        self._precompute()
-
+    
     def _precompute(self):
         L1 = [likelihood.distribution.sqrtprec for likelihood in self.likelihoods]
         L2 = self.prior.sqrtprec
@@ -188,6 +149,11 @@ class LinearRTONew(SamplerNew):
 
         if not hasattr(self.prior, "sqrtprecTimesMean"):
             raise TypeError("Prior must contain a sqrtprecTimesMean attribute")
+    
+    @property
+    def _default_initial_point(self):
+        """ Get the default initial point for the sampler. Defaults to an array of zeros. """
+        return np.zeros(self.dim)
 
 class RegularizedLinearRTONew(LinearRTONew):
     """
@@ -231,15 +197,22 @@ class RegularizedLinearRTONew(LinearRTONew):
         self.stepsize = stepsize
         self.abstol = abstol   
         self.adaptive = adaptive
-        self.proximal = target.prior.proximal
-        self._stepsize = self._choose_stepsize()
         self.maxit = maxit
 
-    @LinearRTONew.target.setter
-    def target(self, value):
-        if not callable(value.prior.proximal):
-            raise TypeError("Projector needs to be callable")
-        return super(RegularizedLinearRTONew, type(self)).target.fset(self, value)
+    def _initialize(self):
+        super()._initialize()
+        self._stepsize = self._choose_stepsize()
+
+    @property
+    def proximal(self):
+        return self.target.prior.proximal
+    
+    def validate_target(self):
+        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")
 
     def _choose_stepsize(self):
         if isinstance(self.stepsize, str):