Changeset aa25fc7 in sasmodels

Timestamp:

May 23, 2018 3:55:27 PM (7 years ago)

Author:

Paul Kienzle <pkienzle@…>

Branches:

master

Children:

910c0f4

Parents:

33969b6

Message:

load user-defined weight functions from ~/.sasview/weights/*.py

Files:

• doc/guide/pd/polydispersity.rst (modified) (3 diffs)
• sasmodels/compare.py (modified) (3 diffs)
• sasmodels/sasview_model.py (modified) (1 diff)
• sasmodels/weights.py (modified) (1 diff)

doc/guide/pd/polydispersity.rst

@@ -28 +28 @@
 sigmas $N_\sigma$ to include from the tails of the distribution, and the
 number of points used to compute the average. The center of the distribution
-is set by the value of the model parameter. The meaning of a polydispersity 
-parameter *PD* (not to be confused with a molecular weight distributions 
-in polymer science) in a model depends on the type of parameter it is being 
+is set by the value of the model parameter. The meaning of a polydispersity
+parameter *PD* (not to be confused with a molecular weight distributions
+in polymer science) in a model depends on the type of parameter it is being
 applied too.
 
-The distribution width applied to *volume* (ie, shape-describing) parameters 
-is relative to the center value such that $\sigma = \mathrm{PD} \cdot \bar x$. 
-However, the distribution width applied to *orientation* (ie, angle-describing) 
+The distribution width applied to *volume* (ie, shape-describing) parameters
+is relative to the center value such that $\sigma = \mathrm{PD} \cdot \bar x$.
+However, the distribution width applied to *orientation* (ie, angle-describing)
 parameters is just $\sigma = \mathrm{PD}$.
 
@@ -73 +73 @@
 or angular orientations, use the Gaussian or Boltzmann distributions.
 
-If applying polydispersion to parameters describing angles, use the Uniform 
-distribution. Beware of using distributions that are always positive (eg, the 
+If applying polydispersion to parameters describing angles, use the Uniform
+distribution. Beware of using distributions that are always positive (eg, the
 Lognormal) because angles can be negative!
 
@@ -315 +315 @@
 .. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
 
+User-defined Distributions
+^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+You can define your own distribution by creating a python file defining a
+*Distribution* object.  The distribution is parameterized by *center*
+(which is always zero for orientation dispersity, or parameter value for
+size dispersity), *sigma* (which is the distribution width in degrees for
+orientation parameters, or center times width for size dispersity), and
+bounds *lb* and *ub* (which are the bounds on the possible values of the
+parameter given in the model definition).
+
+For example, the following wraps the Laplace distribution from scipy stats::
+
+    import numpy as np
+    from scipy.stats import laplace
+
+    from sasmodels import weights
+
+    class Dispersion(weights.Dispersion):
+        r"""
+        Laplace distribution
+
+        .. math::
+
+            w(x) = e^{-\sigma |x - \mu|}
+        """
+        type = "laplace"
+        default = dict(npts=35, width=0, nsigmas=3)  # default values
+        def _weights(self, center, sigma, lb, ub):
+            x = self._linspace(center, sigma, lb, ub)
+            wx = laplace.pdf(x, center, sigma)
+            return x, wx
+
+To see that the distribution is correct use the following::
+
+    from numpy import inf
+    from matplotlib import pyplot as plt
+    from sasmodels import weights
+
+    # reload the user-defined weights
+    weights.load_weights()
+    x, wx = weights.get_weights('laplace', n=35, width=0.1, nsigmas=3, value=50,
+                                limits=[0, inf], relative=True)
+
+    # plot the weights
+    plt.interactive(True)
+    plt.plot(x, wx, 'x')
+
+Any python code can be used to define the distribution.  The distribution
+parameters are available as *self.npts*, *self.width* and *self.nsigmas*.
+Try to follow the convention of gaussian width, npts and number of sigmas
+in the tail, but if your distribution requires more parameters you are free
+to interpret them as something else.  In particular, npts allows you to
+trade accuracy against running time when evaluating your models.  The
+*self._linspace* function uses *self.npts* and *self.nsigmas* to define
+the set of *x* values to use for the distribution (along with the *center*,
+*sigma*, *lb*, and *ub* passed as parameters).  You can use an arbitrary
+set of *x* points.
+
+.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
+
 Note about DLS polydispersity
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

sasmodels/compare.py

@@ -41 +41 @@
 from . import kerneldll
 from . import kernelcl
+from . import weights
 from .data import plot_theory, empty_data1D, empty_data2D, load_data
 from .direct_model import DirectModel, get_mesh
 from .generate import FLOAT_RE, set_integration_size
-from .weights import plot_weights
 
 # pylint: disable=unused-import
@@ -771 +771 @@
             model_info = base._kernel.info
             dim = base._kernel.dim
-            plot_weights(model_info, get_mesh(model_info, base_pars, dim=dim))
+            weights.plot_weights(model_info, get_mesh(model_info, base_pars, dim=dim))
     if opts['plot']:
         import matplotlib.pyplot as plt
@@ -1353 +1353 @@
     #import pprint; pprint.pprint(model_info)
 
+    # Hack to load user-defined distributions; run through all parameters
+    # and make sure any pd_type parameter is a defined distribution.
+    if (any(p.endswith('pd_type') and v not in weights.MODELS
+            for p, v in pars.items()) or
+        any(p.endswith('pd_type') and v not in weights.MODELS
+            for p, v in pars2.items())):
+       weights.load_weights()
+
     if opts['show_pars']:
         if model_info.name != model_info2.name or pars != pars2:

sasmodels/sasview_model.py

@@ -30 +30 @@
 from . import modelinfo
 from .details import make_kernel_args, dispersion_mesh
+
+# Hack: load in any custom distributions
+# Uses ~/.sasview/weights/*.py unless SASMODELS_WEIGHTS is set in the environ.
+# Override with weights.load_weights(pattern="<weights_path>/*.py")
+weights.load_weights()
 
 # pylint: disable=unused-import

sasmodels/weights.py

@@ -231 +231 @@
 ))
 
+SASMODELS_WEIGHTS = "~/.sasview/weights/*.py"
+def load_weights(pattern=None):
+    # type: (str) -> None
+    """
+    Load dispersion distributions matching the given pattern
+    """
+    import logging
+    import os
+    import os.path
+    import glob
+    import traceback
+    from .custom import load_custom_kernel_module
+    if pattern is None:
+        pattern = os.environ.get("SASMODELS_WEIGHTS", SASMODELS_WEIGHTS)
+    for filename in sorted(glob.glob(os.path.expanduser(pattern))):
+        try:
+            print("loading weights from", filename)
+            module = load_custom_kernel_module(filename)
+            MODELS[module.Dispersion.type] = module.Dispersion
+        except Exception as exc:
+            logging.error(traceback.format_exc(exc))
 
 def get_weights(disperser, n, width, nsigmas, value, limits, relative):