Merge pull request #271 from ICB-DCM/develop

Release 0.10.0

.gitignore

@@ -89,3 +89,5 @@ doc/*.log
 .pytest_cache
 dask-worker-space
 *.lock
+*tmp*
+*amici_models*

.travis.yml

@@ -7,6 +7,8 @@ os:
 dist:
 # take the latest supported LTS version
 - xenial
+compiler:
+- gcc
 
 # install dependencies
 
@@ -16,23 +18,29 @@ before_install:
 - sudo add-apt-repository 'deb https://cran.rstudio.com/bin/linux/ubuntu xenial/'
 - sudo apt-get update
 - sudo apt-get install r-base r-base-dev r-base-core r-recommended
-# from apt
+
 addons:
   apt:
     update: true
     packages:
     - redis-server
-# from pip
+    - libhdf5-serial-dev
+    - libatlas-base-dev
+    - swig3.0
+
 install:
+- mkdir -p ~/bin/ && ln -s /usr/bin/swig3.0 ~/bin/swig && export PATH=~/bin/:$PATH
 - pip install --upgrade .
 - pip install --upgrade -r .travis_pip_reqs.txt
 - pip install --upgrade pytest
+- pip install https://github.com/icb-dcm/petab/archive/develop.zip
+- pip install -e git+https://github.com/icb-dcm/amici.git@develop#egg=amici\&subdirectory=python/sdist
 
 # run tests
 
 script:
-- travis_wait 20 python -m pytest --cov=pyabc test/test_*
-- travis_wait 5 xvfb-run -a python -m pytest --cov=pyabc --cov-append test/visualization/test_*
+- travis_wait 10 python -m pytest --cov=pyabc test/test_*
+- travis_wait 3 xvfb-run -a python -m pytest --cov=pyabc --cov-append test/visualization/test_*
 - if [ "$TRAVIS_PULL_REQUEST" != "false" ]; then travis_wait 15 test/run_notebooks.sh; fi
 
 after_success:

doc/examples.rst

@@ -22,6 +22,7 @@ The following examples should help to get a better idea of how to use pyABC.
    examples/external_simulators.ipynb
    examples/data_plots.ipynb
    examples/noise.ipynb
+   examples/petab.ipynb
 
 
 Download the examples as notebooks
@@ -42,6 +43,7 @@ Download the examples as notebooks
 * :download:`External Simulators <examples/external_simulators.ipynb>`
 * :download:`Data plots <examples/data_plots.ipynb>`
 * :download:`Measurement noise assessment <examples/noise.ipynb>`
+* :download:`PEtab import <examples/petab.ipynb>`
 
 
 .. warning::

doc/releasenotes.rst

@@ -4,6 +4,23 @@ Release Notes
 =============
 
 
+0.10 series
+...........
+
+
+0.10.0 (2020-02-20)
+-------------------
+
+* Exact inference via stochastic acceptor finalized and tested (developed
+  throughout the 0.9 series).
+* Support basic PEtab functionality using AMICI ODE simulations (#268).
+* Various error fixes (#265, #267).
+* Log number of processes used by multiprocessing samplers (#263).
+* Implement pyabc.acceptor.ScaledPDFNorm (#269).
+* Implement list population size (#274, #276).
+* On history loading, automatically find an id of a successful run (#273).
+
+
 0.9 series
 ..........
 

flake8_exclude.txt

@@ -1 +1 @@
-doc/conf.py,test_nondeterministic,build,test_performance,tmp
+doc/conf.py,test_nondeterministic,build,test_performance,tmp,*amici_models*

pyabc/__init__.py

@@ -33,6 +33,7 @@
     PercentileDistance,
     RangeEstimatorDistance,
     DistanceWithMeasureList,
+    StochasticKernel,
     NormalKernel,
     IndependentNormalKernel,
     IndependentLaplaceKernel,
@@ -75,7 +76,8 @@
     UniformAcceptor,
     StochasticAcceptor,
     pdf_norm_from_kernel,
-    pdf_norm_max_found)
+    pdf_norm_max_found,
+    ScaledPDFNorm)
 from .model import (
     Model,
     SimpleModel,
@@ -115,6 +117,7 @@
     "PercentileDistance",
     "RangeEstimatorDistance",
     "DistanceWithMeasureList",
+    "StochasticKernel",
     "NormalKernel",
     "IndependentNormalKernel",
     "IndependentLaplaceKernel",
@@ -174,6 +177,7 @@
     "StochasticAcceptor",
     "pdf_norm_from_kernel",
     "pdf_norm_max_found",
+    "ScaledPDFNorm",
     # model
     "ModelResult",
     "Model",

pyabc/acceptor/__init__.py

@@ -17,6 +17,7 @@
 from .pdf_norm import (
     pdf_norm_from_kernel,
     pdf_norm_max_found,
+    ScaledPDFNorm,
 )
 
 
@@ -30,4 +31,5 @@
     # pdf norm
     'pdf_norm_from_kernel',
     'pdf_norm_max_found',
+    'ScaledPDFNorm',
 ]

pyabc/acceptor/pdf_norm.py

@@ -1,5 +1,6 @@
 import numpy as np
-from typing import Callable
+from typing import Callable, Union
+import pandas as pd
 
 
 def pdf_norm_from_kernel(
@@ -13,8 +14,8 @@ def pdf_norm_from_kernel(
 
 
 def pdf_norm_max_found(
-        prev_pdf_norm: float,
-        get_weighted_distances: Callable,
+        prev_pdf_norm: Union[float, None],
+        get_weighted_distances: Callable[[], pd.DataFrame],
         **kwargs):
     """
     Take as pdf_max the maximum over the values found so far in the history,
@@ -34,3 +35,76 @@ def pdf_norm_max_found(
     pdf_norm = max(prev_pdf_norm, *pdfs)
 
     return pdf_norm
+
+
+class ScaledPDFNorm:
+    """
+    Finds the previously found maximum density value, but then scales it
+    by a factor `factor**T` such that the probability of particles getting
+    accepted is increased by y value of up to `factor`.
+
+    Some additional rules are applied to make the scheme stable. The scaling
+    is in particular applied only after a minimum acceptance rate has been
+    violated.
+
+    Parameters
+    ----------
+    factor:
+        The factor by which to effectively rescale the acceptance step's
+        normalization constant.
+    alpha:
+        The ratio by which the subsequent temperature is assumed to be
+        reduced relative to the current one. This is only accurate if a
+        pyabc.ExponentialDecayFixedRatioScheme with corresponding ratio
+        is employed.
+    min_acceptance_rate:
+        The scaling is applied once the acceptance rates fall below this
+        value for the first time.
+    """
+
+    def __init__(
+            self,
+            factor: float = 10,
+            alpha: float = 0.5,
+            min_acceptance_rate: bool = 0.1):
+        self.factor = 10
+        self.alpha = alpha
+        self.min_acceptance_rate = min_acceptance_rate
+        self._hit = False
+
+    def __call__(
+            self,
+            prev_pdf_norm: Union[float, None],
+            get_weighted_distances: Callable[[], pd.DataFrame],
+            prev_temp: Union[float, None],
+            acceptance_rate: float,
+            **kwargs):
+        # base: the maximum found temperature
+        pdf_norm = pdf_norm_max_found(
+            prev_pdf_norm=prev_pdf_norm,
+            get_weighted_distances=get_weighted_distances)
+
+        # log-scale
+        offset = np.log(self.factor)
+
+        if acceptance_rate >= self.min_acceptance_rate and not self._hit:
+            # do not apply scaling yet since acceptance rates still feasible
+            return pdf_norm
+
+        # from now on rescale
+        self._hit = True
+
+        if prev_temp is None:
+            # can't take temperature into account, thus effectively assume T=1
+            next_temp = 1
+        else:
+            # note: this is only accurate if the temperature is based on a
+            #  ExponentialDecayFixedRatioScheme with the given alpha value
+            next_temp = self.alpha * prev_temp
+
+        # the offset is multiplied by the next temperature so that the
+        #  effective resulting factor in the acceptance step is as intended
+        scaled_norm = pdf_norm - offset * next_temp
+
+        # used_norm = max(prev_pdf_norm, used_norm)
+        return scaled_norm

pyabc/parameters.py

@@ -1,7 +1,4 @@
-from collections import UserDict
-
-
-class ParameterStructure(UserDict):
+class ParameterStructure(dict):
     """
     Basic functionality of a structure containing parameters.
     """

pyabc/petab/__init__.py

@@ -0,0 +1,6 @@
+from .amici import AmiciPetabImporter
+
+
+__all__ = [
+    'AmiciPetabImporter'
+]