CUQIpy 1.1.0.post0.dev32__tar.gz → 1.1.0.post0.dev39__tar.gz

{cuqipy-1.1.0.post0.dev32 → cuqipy-1.1.0.post0.dev39}/CUQIpy.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: CUQIpy
-Version: 1.1.0.post0.dev32
+Version: 1.1.0.post0.dev39
 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-1.1.0.post0.dev32 → cuqipy-1.1.0.post0.dev39}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: CUQIpy
-Version: 1.1.0.post0.dev32
+Version: 1.1.0.post0.dev39
 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-1.1.0.post0.dev32 → cuqipy-1.1.0.post0.dev39}/cuqi/_version.py

@@ -8,11 +8,11 @@ import json
 
 version_json = '''
 {
- "date": "2024-08-26T11:16:28+0200",
+ "date": "2024-08-29T10:46:43+0300",
  "dirty": false,
  "error": null,
- "full-revisionid": "0a3eb53894cc988444a61e5b895cee57b66747a9",
- "version": "1.1.0.post0.dev32"
+ "full-revisionid": "9d8920342c73b1bd8e7a47a216eb34056ea2f5d6",
+ "version": "1.1.0.post0.dev39"
 }
 '''  # END VERSION_JSON
 

{cuqipy-1.1.0.post0.dev32 → cuqipy-1.1.0.post0.dev39}/cuqi/utilities/__init__.py

@@ -11,6 +11,8 @@ from ._utilities import (
     sparse_cholesky,
     approx_derivative,
     check_if_conditional_from_attr,
+    plot_1D_density,
+    plot_2D_density
 )
 
 from ._get_python_variable_name import _get_python_variable_name

{cuqipy-1.1.0.post0.dev32 → cuqipy-1.1.0.post0.dev39}/cuqi/utilities/_utilities.py

@@ -1,4 +1,5 @@
 from cuqi.array import CUQIarray
+from cuqi.density import Density
 import numpy as np
 import inspect
 from numbers import Number
@@ -7,6 +8,7 @@ from scipy.sparse import linalg as spslinalg
 from dataclasses import dataclass
 from abc import ABCMeta
 import copy
+import matplotlib.pyplot as plt
 
 
 def force_ndarray(value,flatten=False):
@@ -229,3 +231,117 @@ def approx_gradient(func, x, epsilon= 0.000001):
         eps_vec[i] = 0.0
         
     return FD_gradient
+
+
+# Function for plotting 1D density functions
+def plot_1D_density(density:Density,
+                    v_min, v_max,
+                    N=501, log_scale=False,
+                    **kwargs):
+    """ Plot 1D density function 
+
+    Parameters
+    ----------
+    density : CUQIpy Density
+        The density to be plotted.
+
+    v_min : float
+        Minimum value for the variable.
+    
+    v_max : float
+        Maximum value for the variable.
+
+    N : int
+        Number of grid points for the variable.
+
+    log_scale : bool
+        If True, the density is plotted in log scale.
+    
+    kwargs : dict
+        Additional keyword arguments for the plot that are passed to the
+        underlying plotting method: `matplotlib.pyplot.plot` function 
+        in this case.
+
+    """
+    # Assert that the density is 1D
+    assert density.dim == 1, "The density must be for a scalar variable"
+    ls = np.linspace(v_min, v_max, N)
+
+    # Create a map to evaluate density
+    density_map = (lambda x: x) if log_scale else (lambda x: np.exp(x))
+
+    # Evaluate density on grid
+    y = [density_map(density.logd(grid_point)) for grid_point in ls]
+    p = plt.plot(ls, y, **kwargs)
+    return p
+
+# Function for plotting 2D density functions
+def plot_2D_density(density: Density, 
+                    v1_min, v1_max,
+                    v2_min, v2_max,
+                    N1=201, N2=201,
+                    log_scale=False,
+                    **kwargs):
+    """ Plot 2D density function 
+
+    Parameters
+    ----------
+    density : CUQIpy Density
+        The density to be plotted.
+
+    v1_min : float
+        Minimum value for the first variable.
+    
+    v1_max : float
+        Maximum value for the first variable.
+    
+    v2_min : float
+        Minimum value for the second variable.
+
+    v2_max : float  
+        Maximum value for the second variable.
+
+    N1 : int
+        Number of grid points for the first variable.
+
+    N2 : int
+        Number of grid points for the second variable.
+    
+    log_scale : bool
+        If True, the density is plotted in log scale.
+
+    kwargs : dict
+        Additional keyword arguments for the plot that are passed to the
+        underlying plotting method: `matplotlib.pyplot.imshow` function 
+        in this case.
+
+    """
+    # Assert that the density is 2D
+    assert density.dim == 2,\
+        "The density must be for a two-dimensional variable"
+    # Create grid
+    ls1 = np.linspace(v1_min, v1_max, N1)
+    ls2 = np.linspace(v2_min, v2_max, N2)
+    grid1, grid2 = np.meshgrid(ls1, ls2)
+
+    # Create a map to evaluate density
+    density_map = (lambda x: x) if log_scale else (lambda x: np.exp(x))
+
+    # Evaluate density on grid
+    evaluated_density = np.zeros((N1, N2))
+    for ii in range(N1):
+        for jj in range(N2):
+            evaluated_density[ii,jj] = density_map(
+                density.logd([grid1[ii,jj], grid2[ii,jj]])) 
+
+    # Plot
+    pixelwidth_x = (v1_max-v1_min)/(N2-1)
+    pixelwidth_y = (v2_max-v2_min)/(N2-1)
+
+    hp_x = 0.5*pixelwidth_x
+    hp_y = 0.5*pixelwidth_y
+
+    extent = (v1_min-hp_x, v1_max+hp_x, v2_min-hp_y, v2_max+hp_y)
+
+    im = plt.imshow(evaluated_density, origin='lower', extent=extent, **kwargs)
+    return im

cuqipy-1.1.0.post0.dev39/tests/test_utilities.py

@@ -0,0 +1,68 @@
+import pytest
+from scipy.linalg import cholesky
+from scipy.sparse import diags
+from cuqi.utilities import sparse_cholesky, plot_1D_density, plot_2D_density
+from cuqi.model import LinearModel
+from cuqi.distribution import Gaussian, Uniform, JointDistribution
+import numpy as np
+import matplotlib.pyplot as plt
+
+
+@pytest.mark.parametrize("P", [
+    diags([-1, 2, -1], [-1, 0, 1], shape=(128, 128)),
+    diags([1, -4, 6, -4, 1], [-2, -1, 0, 1, 2], shape=(128, 128))
+])
+def test_sparse_cholesky(P):
+    """ Test the sparse Cholesky decomposition. P is a sparse matrix (often precision). """
+    # Scipy version (on dense only)
+    L1 = cholesky(P.toarray())
+
+    # Scipy-based version from CUQIpy (on sparse)
+    L2 = sparse_cholesky(P) 
+
+    assert np.allclose(L1, L2.toarray()) # Convert to dense to compare
+
+
+def posterior_2D_components():
+    """A function to create and return 2D Bayesian posterior components"""
+    # Create a CUQIpy model
+    A = LinearModel(np.array([[1.0, 1.0]]))
+    # Create a CUQIpy prior
+    x = Gaussian(mean=np.array([0.0, 0.0]), cov=np.array([1.0, 1.0]))
+    # Create a CUQIpy likelihood
+    y = Gaussian(A@x, cov=0.1)
+    return x, y
+
+
+def likelihood_2D():
+    """A function to create and return a 2D likelihood"""
+    x, y = posterior_2D_components()
+    likelihood = y.to_likelihood(2)
+    return likelihood
+
+
+def posterior_2D():
+    """A function to create and return a 2D posterior"""
+    x, y = posterior_2D_components()
+    joint = JointDistribution(x, y)
+    posterior = joint(y=2)
+    return posterior
+
+
+@pytest.mark.parametrize("density", [
+    Gaussian(mean=0.0, cov=1.0),
+    Uniform(low=0.0, high=1.0),
+    Gaussian(mean=np.array([0.0, 0.0]), cov=np.array([1.0, 1.0])),
+    likelihood_2D(),
+    posterior_2D()
+])
+@pytest.mark.parametrize("log_scale", [True, False])
+def test_plot_density(density, log_scale):
+    """ Test the 1D and 2D density plot. """
+    if density.dim == 1:
+        plot_1D_density(density, -1, 1, color='green', log_scale=log_scale)
+    elif density.dim == 2:
+        plot_2D_density(
+            density, -3, 3, -3, 3, 60, 60, cmap='gray', log_scale=log_scale)
+    else:
+        raise ValueError("Density must be 1D or 2D.")

cuqipy-1.1.0.post0.dev32/tests/test_utilities.py

@@ -1,20 +0,0 @@
-import pytest
-from scipy.linalg import cholesky
-from scipy.sparse import diags
-from cuqi.utilities import sparse_cholesky
-import numpy as np
-
-
-@pytest.mark.parametrize("P", [
-    diags([-1, 2, -1], [-1, 0, 1], shape=(128, 128)),
-    diags([1, -4, 6, -4, 1], [-2, -1, 0, 1, 2], shape=(128, 128))
-])
-def test_sparse_cholesky(P):
-    """ Test the sparse Cholesky decomposition. P is a sparse matrix (often precision). """
-    # Scipy version (on dense only)
-    L1 = cholesky(P.toarray())
-
-    # Scipy-based version from CUQIpy (on sparse)
-    L2 = sparse_cholesky(P) 
-
-    assert np.allclose(L1, L2.toarray()) # Convert to dense to compare