V 021 (#455)

* Parameters plot, #418 (#425) * parameters plot, plotting a subset of parameters (#418) * "Variable used before assignment" fixed for parameters plot (#364) * Feature autocorrelation - close #251 (#426) * autocorrelation calculation by sokal * docs * notebook using ess * added ess and ac in result * created ess calculation * rename variable * docs * fix * remove print from test * parameter index in 1d marginal plot * added type * added type * docu * flake8 * Update pypesto/sample/diagnostics.py Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * docs * docs * grammar thingy * handle in get_data_to_plot in visu Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * fix display for least squares solver (#430) fix #429 * basic sanity checking for parameter fixing (#420) * basic sanity checking for parameter fixing * fixup type checks * fixup * fix typehints * fixup & refactor * update doc * fixup * Fix amici error output, fixes #427 (#428) * initial fix * add filter function * unify error and base initialization of return values * Profile options (#419) * change profile indexing to parameter indices instead of boolean indexing * adapt also approximation and test to new profile api Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Prior Example, closes # 434 (#438) * Add notebook prior definition * fix sphinx * Update doc/example/prior_definition.ipynb Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Update doc/example/prior_definition.ipynb Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Update doc/example/prior_definition.ipynb Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Update doc/example/prior_definition.ipynb Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Update doc/example/prior_definition.ipynb Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Update doc/example/prior_definition.ipynb Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Update doc/example/prior_definition.ipynb Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Update doc/example/prior_definition.ipynb Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Update doc/example/prior_definition.ipynb Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Improve notebook Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Fix offset_y and scale_y in waterfal plot (#441) * Correct offset_y for waterfall plots of result lists * enforce y_limits which show whole data when plotting result lists, if no explicit user input was given * fix nans and infs, deal with possibly different number of optiimzation runs in different result objects * fix 2 typos (#444) Co-authored-by: Thomas S. Ligon <tom@thomassligon.info> * Update priors.py (#448) * Fix optimization with priors (#452) * Update priors.py * Update priors.py * Fix PEtab import (Priors) (#454), closes #447 * bugfix + compliance to current PEtab version in prior import * resolve review + flake8 Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * prepare version 0.2.1 * Update doc/releasenotes.rst Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Update doc/releasenotes.rst Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> * Update doc/releasenotes.rst Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> Co-authored-by: Polina Lakrisenko <p.lakrisenko@gmail.com> Co-authored-by: Elba Raimúndez Alvarez <elba.raimundez@helmholtz-muenchen.de> Co-authored-by: Yannik Schälte <31767307+yannikschaelte@users.noreply.github.com> Co-authored-by: Fabian Fröhlich <fabian@schaluck.com> Co-authored-by: Paul Stapor <paul.stapor@helmholtz-muenchen.de> Co-authored-by: Thomas S. Ligon <tom@thomassligon.info>
ICB-DCM · Sep 7, 2020 · cbec7ee · cbec7ee
1 parent 8951c39
commit cbec7ee
Show file tree

Hide file tree

Showing 29 changed files with 1,195 additions and 336 deletions.
diff --git a/doc/example.rst b/doc/example.rst
@@ -14,6 +14,7 @@ The following examples cover typical use cases and should help get a better idea
    example/store.ipynb
    example/sampler_study.ipynb
    example/sampling_diagnostics.ipynb
+   example/prior_definition.ipynb
 
 Download the examples as notebooks
 ----------------------------------
@@ -26,6 +27,7 @@ Download the examples as notebooks
 * :download:`Storage <example/store.ipynb>`
 * :download:`Sampler study <example/sampler_study.ipynb>`
 * :download:`Sampling diagnostics <example/sampling_diagnostics.ipynb>`
+* :download:`Prior definition <example/prior_definition.ipynb>`
 
 .. Note::
    Some of the notebooks have extra dependencies.
diff --git a/doc/example/fixed_parameters.ipynb b/doc/example/fixed_parameters.ipynb
diff --git a/doc/example/prior_definition.ipynb b/doc/example/prior_definition.ipynb
diff --git a/doc/example/sampling_diagnostics.ipynb b/doc/example/sampling_diagnostics.ipynb
diff --git a/doc/releasenotes.rst b/doc/releasenotes.rst
@@ -5,6 +5,20 @@ Release notes
 0.2 series
 ..........
 
+0.2.1 (2020-09-07)
+------------------
+
+* Example Notebook for prior functionality (#438)
+* Changed parameter indexing in profiling routines (#419)
+* Basic sanity checking for parameter fixing (#420)
+* Bug fixes in:
+  * Displaying of multi start optimization (#430)
+  * AMICI error output (#428)
+  * Axes scaling/limits in waterfall plots (#441)
+  * Priors (PEtab impor, error handling) (#448, #452, #454)
+* Improved sampling diagnostics (e.g. effective samples size) (#426)
+* Improvements and bug fixes in parameter plots (#425)
+
 
 0.2.0 (2020-06-17)
 ------------------

diff --git a/pypesto/objective/amici.py b/pypesto/objective/amici.py
@@ -148,7 +148,7 @@ def __init__(self,
         # need to do preequilibration
         if self.guess_steadystate:
             if self.amici_model.ncl() > 0:
-                raise ValueError('Steadystate prediciton is not supported for'
+                raise ValueError('Steadystate prediction is not supported for '
                                  'models with conservation laws!')
 
             if self.amici_model.getSteadyStateSensitivityMode() == \

diff --git a/pypesto/objective/amici_calculator.py b/pypesto/objective/amici_calculator.py
@@ -6,7 +6,9 @@
 )
 from .amici_util import (
     add_sim_grad_to_opt_grad, add_sim_hess_to_opt_hess,
-    sim_sres_to_opt_sres, log_simulation, get_error_output)
+    sim_sres_to_opt_sres, log_simulation, get_error_output, filter_return_dict,
+    init_return_values
+)
 
 try:
     import amici
@@ -128,24 +130,11 @@ def calculate_function_values(rdatas,
 
     # check if the simulation failed
     if any(rdata['status'] < 0.0 for rdata in rdatas):
-        return get_error_output(amici_model, edatas, rdatas, dim)
-
-    # prepare outputs
-    nllh = 0.0
-    snllh = None
-    s2nllh = None
-    if mode == MODE_FUN and sensi_order > 0:
-        snllh = np.zeros(dim)
-        s2nllh = np.zeros([dim, dim])
-
-    chi2 = None
-    res = None
-    sres = None
-    if mode == MODE_RES:
-        chi2 = 0.0
-        res = np.zeros([0])
-        if sensi_order > 0:
-            sres = np.zeros([0, dim])
+        return get_error_output(amici_model, edatas, rdatas,
+                                sensi_order, mode, dim)
+
+    nllh, snllh, s2nllh, chi2, res, sres = init_return_values(sensi_order,
+                                                              mode, dim)
 
     par_sim_ids = list(amici_model.getParameterIds())
     sensi_method = amici_solver.getSensitivityMethod()
@@ -212,8 +201,4 @@ def calculate_function_values(rdatas,
         SRES: sres,
         RDATAS: rdatas
     }
-    return {
-        key: val
-        for key, val in ret.items()
-        if val is not None
-    }
+    return filter_return_dict(ret)
diff --git a/pypesto/objective/amici_util.py b/pypesto/objective/amici_util.py
@@ -3,7 +3,9 @@
 from typing import Dict, Sequence, Union
 import logging
 
-from .constants import FVAL, GRAD, HESS, RES, SRES, RDATAS
+from .constants import (
+    FVAL, CHI2, GRAD, HESS, RES, SRES, RDATAS, MODE_FUN, MODE_RES
+)
 
 try:
     import amici
@@ -224,6 +226,8 @@ def get_error_output(
         amici_model: AmiciModel,
         edatas: Sequence['amici.ExpData'],
         rdatas: Sequence['amici.ReturnData'],
+        sensi_order: int,
+        mode: str,
         dim: int):
     """Default output upon error.
 
@@ -237,11 +241,56 @@ def get_error_output(
                  for data in edatas)
     n_res = nt * amici_model.nytrue
 
-    return {
-        FVAL: np.inf,
-        GRAD: np.nan * np.ones(dim),
-        HESS: np.nan * np.ones([dim, dim]),
-        RES:  np.nan * np.ones(n_res),
-        SRES: np.nan * np.ones([n_res, dim]),
+    nllh, snllh, s2nllh, chi2, res, sres = init_return_values(sensi_order,
+                                                              mode, dim, True)
+    if res is not None:
+        res = np.nan * np.ones(n_res)
+    if sres is not None:
+        sres = np.nan * np.ones([n_res, dim])
+
+    ret = {
+        FVAL: nllh,
+        CHI2: chi2,
+        GRAD: snllh,
+        HESS: s2nllh,
+        RES: res,
+        SRES: sres,
         RDATAS: rdatas
     }
+    return filter_return_dict(ret)
+
+
+def init_return_values(sensi_order, mode, dim, error=False):
+    if error:
+        fval = np.inf
+        sval = np.nan
+    else:
+        fval = sval = 0.0
+
+    nllh = fval
+    snllh = None
+    s2nllh = None
+    if mode == MODE_FUN and sensi_order > 0:
+        snllh = sval * np.ones(dim)
+        if sensi_order > 1:
+            s2nllh = sval * np.ones([dim, dim])
+
+    chi2 = None
+    res = None
+    sres = None
+    if mode == MODE_RES:
+        chi2 = fval
+        res = np.zeros([0])
+        if sensi_order > 0:
+            sres = np.zeros([0, dim])
+
+    return nllh, snllh, s2nllh, chi2, res, sres
+
+
+def filter_return_dict(ret):
+    """Filters return dict for non-None values"""
+    return {
+        key: val
+        for key, val in ret.items()
+        if val is not None
+    }
diff --git a/pypesto/objective/priors.py b/pypesto/objective/priors.py
@@ -78,15 +78,17 @@ def call_unprocessed(
 
         res = {}
 
+        res[FVAL] = self.neg_log_density(x)
+
         for order in sensi_orders:
             if order == 0:
-                res[FVAL] = self.neg_log_density(x)
+                continue
             elif order == 1:
                 res[GRAD] = self.gradient_neg_log_density(x)
             elif order == 2:
                 res[HESS] = self.hessian_neg_log_density(x)
             else:
-                ValueError(f'Invalid sensi order {order}.')
+                raise ValueError(f'Invalid sensi order {order}.')
 
         return res
 
@@ -104,8 +106,8 @@ def check_mode(self, mode) -> bool:
         elif mode == MODE_RES:
             return False
         else:
-            ValueError(f'Invalid input: Expected mode {MODE_FUN} or'
-                       f' {MODE_RES}, received {mode} instead.')
+            raise ValueError(f'Invalid input: Expected mode {MODE_FUN} or'
+                             f' {MODE_RES}, received {mode} instead.')
 
     def neg_log_density(self, x):
         """
@@ -347,8 +349,8 @@ def dd_log_f_ddx(x):
         # when implementing: add to tests
         raise NotImplementedError
     else:
-        ValueError(f'NegLogPriors of type {prior_type} are currently '
-                   'not supported')
+        raise ValueError(f'NegLogPriors of type {prior_type} are currently '
+                         'not supported')
 
 
 def _get_linear_function(slope: float,

diff --git a/pypesto/optimize/optimizer.py b/pypesto/optimize/optimizer.py
@@ -307,9 +307,12 @@ def minimize(
             # TODO: pass jac computing methods in options
 
             if self.options is not None:
-                self.options['verbose'] = 2 if 'disp' in self.options.keys() \
-                                               and self.options['disp'] else 0
-                self.options.pop('disp', None)
+                ls_options = self.options.copy()
+                ls_options['verbose'] = 2 if 'disp' in ls_options.keys() \
+                                             and ls_options['disp'] else 0
+                ls_options.pop('disp', None)
+            else:
+                ls_options = {}
 
             # optimize
             res = scipy.optimize.least_squares(
@@ -320,7 +323,7 @@ def minimize(
                 bounds=bounds,
                 tr_solver='exact',
                 loss='linear',
-                **self.options
+                **ls_options
             )
             # extract fval/grad from result, note that fval is not available
             # from least squares solvers

diff --git a/pypesto/petab/importer.py b/pypesto/petab/importer.py
@@ -1,5 +1,4 @@
 import pandas as pd
-import numpy as np
 import os
 import sys
 import importlib
@@ -283,10 +282,8 @@ def create_prior(self) -> NegLogParameterPriors:
                 prior_type_entry = self.petab_problem.\
                     parameter_df.loc[x_id, petab.OBJECTIVE_PRIOR_TYPE]
 
-                # TODO: Change the hardcoded "uninformative" to
-                #  petab.UNINFORMATIVE, if corresponding PEtab PR is merged
-                if not (np.isnan(prior_type_entry)
-                        or prior_type_entry == 'uninformative'):
+                if (isinstance(prior_type_entry, str)
+                        and prior_type_entry != petab.PARAMETER_SCALE_UNIFORM):
 
                     prior_params = [float(param) for param in
                                     self.petab_problem.parameter_df.