Changeset 74b0495 in sasmodels

Timestamp:

Oct 10, 2017 1:39:39 PM (7 years ago)

Author:

Paul Kienzle <pkienzle@…>

Branches:

master, core_shell_microgels, costrafo411, magnetic_model, ticket-1257-vesicle-product, ticket_1156, ticket_1265_superball, ticket_822_more_unit_tests

Children:

09141ff

Parents:

b76191e

Message:

update oriented sesans and simultaneous fit examples

Files:

• example/latex_smeared_out_0.txt (deleted)
• example/latex_smeared_out_1.txt (deleted)
• example/oriented_usans.py (modified) (1 diff)
• example/simul_fit.py (modified) (2 diffs)
• sasmodels/bumps_model.py (modified) (2 diffs)
• sasmodels/data.py (modified) (3 diffs)

example/oriented_usans.py

@@ -6 +6 @@
 
 # Spherical particle data, not ellipsoids
-sans, usans = load_data('../../sasview/sasview/test/1d_data/latex_smeared.xml')
+sans, usans = load_data('latex_smeared.xml', index='all')
 usans.qmin, usans.qmax = np.min(usans.x), np.max(usans.x)
 usans.mask = (usans.x < 0.0)

example/simul_fit.py

@@ -1 @@
-#!/usr/bin/env python
-# -*- coding: utf-8 -*-
-
-# To Sasview/documents/scripts
-
 from bumps.names import *
 from sasmodels.core import load_model
@@ -9 +4 @@
 from sasmodels.data import load_data, plot_data
 
+# latex data, same sample usans and sans
+# particles radius ~2300, uniform dispersity
+datasets = load_data('latex_smeared.xml', index='all')
+#[print(data) for data in datasets]
 
-""" IMPORT THE DATA USED """
-datafiles = ['latex_smeared_out_0.txt', 'latex_smeared_out_1.txt']
-datasets = [load_data(el) for el in datafiles]
+# A single sphere model to share between the datasets.  We will use
+# FreeVariables below to set the parameters that are independent between
+# the datasets.
+kernel = load_model('sphere')
+pars = dict(scale=0.01, background=0.0, sld=5.0, sld_solvent=0.0, radius=1500.,
+            #radius_pd=0.1, radius_pd_n=35,
+            )
+model = Model(kernel, **pars)
 
-for data in datasets:
-    data.qmin = 0.0
-    data.qmax = 10.0
+# radius and polydispersity (if any) are shared
+model.radius.range(0, inf)
+#model.radius_pd.range(0, 1)
 
-#sphere model
-kernel = load_model('sphere', dtype="single")
-pars = dict(scale=0.01, background=0.0, sld=1.0, sld_solvent=6.0, radius=1500.)
-model = Model(kernel, **pars)
-model.radius.range(0, inf)
-#model.background.range(-inf, inf)
-#model.scale.range(0, inf)
-model.sld.range(-inf, inf)
-model.sld_solvent.range(-inf, inf)
+# Contrast and dilution are the same for both measurements, but are not
+# separable with a single measurement (i.e., I(q) ~ F(q) contrast^2 Vf),
+# so fit one of scale, sld or solvent sld.  With absolute scaling from
+# data reduction, can use the same parameter for both datasets.
+model.scale.range(0, inf)
+#model.sld.range(-inf, inf)
+#model.sld_solvent.range(-inf, inf)
 
+# Background is different for sans and usans so set it as a free variable
+# in the model.
 free = FreeVariables(
-    names=[data.filename for data in datasets],
+    names=[data.run[0] for data in datasets],
     background=model.background,
-    scale=model.scale,
     )
 free.background.range(-inf, inf)
-free.scale.range(0, inf)
 
-M = [Experiment(data=data, model=model) for data in datasets]
+# Note: can access the parameters for the individual models using
+# free.background[0] and free.background[1], setting constraints or
+# ranges as appropriate.
+
+# For more complex systems where different datasets require independent models,
+# separate models can be defined, with parameters tied together using
+# constraint expressions.  For example, the following could be used to fit
+# data set 1 to spheres and data set 2 to cylinders of the same volume:
+#    model1 = Model(load_model('sphere'))
+#    model2 = Model(load_model('cylinder'))
+#    model1.sld = model2.sld
+#    model1.sld_solvent = model2.sld_solvent
+#    model1.scale = model2.scale
+#    # set cylinders and spheres to the same volume
+#    model1.radius = (3/4*model2.radius**2*model2.length)**(1/3)
+#    model1.background.range(0, 2)
+#    model2.background.range(0, 2)
+
+# Setup the experiments, sharing the same model across all datasets.
+M = [Experiment(data=data, model=model, name=data.run[0]) for data in datasets]
 
 problem = FitProblem(M, freevars=free)
-
-print(problem._parameters)

sasmodels/bumps_model.py

@@ -133 +133 @@
     """
     _cache = None # type: Dict[str, np.ndarray]
-    def __init__(self, data, model, cutoff=1e-5):
+    def __init__(self, data, model, cutoff=1e-5, name=None):
         # type: (Data, Model, float) -> None
         # remember inputs so we can inspect from outside
+        self.name = data.filename if name is None else name
         self.model = model
         self.cutoff = cutoff
@@ -204 +205 @@
         """
         data, theory, resid = self._data, self.theory(), self.residuals()
-        plot_theory(data, theory, resid, view, Iq_calc=self.Iq_calc)
+        # TODO: hack to display oriented usans 2-D pattern
+        Iq_calc = self.Iq_calc if isinstance(self.Iq_calc, tuple) else None
+        plot_theory(data, theory, resid, view, Iq_calc=Iq_calc)
 
     def simulate_data(self, noise=None):

sasmodels/data.py

@@ -44 +44 @@
     Data = Union["Data1D", "Data2D", "SesansData"]
 
-def load_data(filename):
+def load_data(filename, index=0):
     # type: (str) -> Data
     """
@@ -55 +55 @@
         filename, indexstr = filename[:-1].split('[')
         index = int(indexstr)
-    else:
-        index = None
     datasets = loader.load(filename)
     if datasets is None:
@@ -62 +60 @@
     if not isinstance(datasets, list):
         datasets = [datasets]
-    if index is None and len(datasets) > 1:
-        raise ValueError("Need to specify filename[index] for multipart data")
-    data = datasets[index if index is not None else 0]
-    if hasattr(data, 'x'):
-        data.qmin, data.qmax = data.x.min(), data.x.max()
-        data.mask = (np.isnan(data.y) if data.y is not None
-                     else np.zeros_like(data.x, dtype='bool'))
-    elif hasattr(data, 'qx_data'):
-        data.mask = ~data.mask
-    return data
+    for data in datasets:
+        if hasattr(data, 'x'):
+            data.qmin, data.qmax = data.x.min(), data.x.max()
+            data.mask = (np.isnan(data.y) if data.y is not None
+                        else np.zeros_like(data.x, dtype='bool'))
+        elif hasattr(data, 'qx_data'):
+            data.mask = ~data.mask
+    return datasets[index] if index != 'all' else datasets