Release 0.10.0

Release 0.10.0

changelog.md

@@ -1,3 +1,17 @@
+# 0.10.0
+
+## Major changes
+
+* ADD further acquisition functions: PI and LCB
+* SMAC can now be installed without installing all its dependencies
+* Simplify setup.py by moving most thing to setup.cfg
+
+## Bug fixes
+
+* RM typing as requirement
+* FIX import of authors in setup.py
+* MAINT use json-file as standard pcs format for internal logging 
+
 # 0.9
 
 ## Major changes

doc/conf.py

@@ -62,8 +62,8 @@
 
 # General information about the project.
 project = u'SMAC'
-copyright = '2015-%s, %s' % (datetime.datetime.now().year, smac.AUTHORS)
-author = smac.AUTHORS
+copyright = '2015-%s, %s' % (datetime.datetime.now().year, smac.__author__)
+author = smac.__author__
 
 # The version info for the project you're documenting, acts as replacement for
 # |version| and |release|, also used in various other places throughout the
@@ -243,7 +243,7 @@
 # (source start file, target name, title,
 #  author, documentclass [howto, manual, or own class]).
 latex_documents = [
-  (master_doc, 'SMAC3.tex', u'SMAC3 Documentation', smac.AUTHORS, 'manual'),
+  (master_doc, 'SMAC3.tex', u'SMAC3 Documentation', smac.__author__, 'manual'),
 ]
 
 # The name of an image file (relative to this directory) to place at the top of

requirements.txt

@@ -7,7 +7,6 @@ psutil
 pynisher>=0.4.1
 ConfigSpace>=0.4.6,<0.5
 scikit-learn>=0.18.0
-typing
 pyrfr>=0.5.0
 sphinx
 sphinx_rtd_theme

setup.cfg

@@ -1,2 +1,24 @@
 [metadata]
-description-file = README.md
+name = smac
+author_email = "[email protected]"
+description = SMAC3, a Python implementation of 'Sequential Model-based Algorithm Configuration'.
+long_description = file: README.md, CHANGELOG.md, LICENSE
+license = 3-clause BSD
+keywords = machine learning, algorithm configuration, hyperparameter optimization, tuning
+classifiers=
+        Development Status :: 3 - Alpha
+        Topic :: Utilities
+        Topic :: Scientific/Engineering
+        Topic :: Scientific/Engineering :: Artificial Intelligence
+        License :: OSI Approved :: BSD License
+platforms = Linux
+
+[options]
+packages = find:
+
+[options.packages.find]
+exclude = test
+
+[options.entry_points]
+console_scripts =
+    smac = smac.smac_cli:cmd_line_call

setup.py

@@ -1,46 +1,37 @@
-import setuptools
-import sys
-
-import smac
+#!/usr/bin/env python3
+import os
+from setuptools import setup
 
 
 with open('requirements.txt') as fh:
     requirements = fh.read()
 requirements = requirements.split('\n')
 requirements = [requirement.strip() for requirement in requirements]
 
-with open("smac/__version__.py") as fh:
-    version = fh.readlines()[-1].split()[-1].strip("\"'")
-
-if sys.version_info < (3, 5, 2):
-    raise ValueError('Unsupported Python version %s found. SMAC3 requires Python 3.5.2 or higher.' % sys.version_info)
-
-setuptools.setup(
-    name="smac",
-    version=version,
-    author=smac.AUTHORS,
-    author_email="[email protected]",
-    description=("SMAC3, a Python implementation of 'Sequential Model-based "
-                 "Algorithm Configuration'."),
-    license="3-clause BSD",
-    keywords="machine learning algorithm configuration hyperparameter "
-             "optimization tuning",
-    url="",
-    entry_points={
-        'console_scripts': ['smac=smac.smac_cli:cmd_line_call'],
-    },
-    packages=setuptools.find_packages(exclude=['test', 'source']),
-    classifiers=[
-        "Development Status :: 3 - Alpha",
-        "Topic :: Utilities",
-        "Topic :: Scientific/Engineering",
-        "Topic :: Scientific/Engineering :: Artificial Intelligence",
-        "License :: OSI Approved :: BSD License",
-    ],
-    platforms=['Linux'],
+
+def get_version():
+    version_file = os.path.join(os.path.dirname(os.path.abspath(__file__)), "smac", "__init__.py")
+    for line in open(version_file):
+        if line.startswith("__version__"):
+            version = line.split("=")[1].strip().replace("'", "").replace('"', '')
+            return version
+    raise RuntimeError("Unable to find version string in %s" % version_file)
+
+
+def get_author():
+    version_file = os.path.join(os.path.dirname(os.path.abspath(__file__)), "smac", "__init__.py")
+    for line in open(version_file):
+        if line.startswith("__author__"):
+            version = line.split("=")[1].strip().replace("'", "").replace('"', '')
+            return version
+    raise RuntimeError("Unable to find author string in %s" % version_file)
+
+
+setup(
+    python_requires=">=3.5.2",
     install_requires=requirements,
-    python_requires='>=3.5.2',
-    tests_require=['mock',
-                   'nose'],
-    test_suite='nose.collector'
+    author=get_author(),
+    version=get_version(),
+    test_suite="nose.collector",
+    tests_require=["mock", "nose"]
 )

smac/__init__.py

@@ -1,33 +1,27 @@
 import os
 import sys
 
-if sys.version_info < (3, 5, 2):
-    raise ValueError("SMAC requires Python 3.5.2 or newer.")
-
-from smac.__version__ import __version__
 from smac.utils import dependencies
 
-AUTHORS = "Marius Lindauer, Matthias Feurer, Katharina Eggensperger, " \
-          "Aaron Klein, Stefan Falkner and Frank Hutter"
+__version__ = '0.10.0'
+__author__ = 'Marius Lindauer, Matthias Feurer, Katharina Eggensperger, Joshua Marben, André Biedenkapp, Aaron Klein, Stefan Falkner and Frank Hutter'
 
-__MANDATORY_PACKAGES__ = '''
+__MANDATORY_PACKAGES__ = """
 numpy>=1.7.1
 scipy>=0.18.1
+six
+psutil
+pynisher>=0.4.1
+ConfigSpace>=0.4.6,<0.5
 scikit-learn>=0.18.0
 pyrfr>=0.5.0
-ConfigSpace>=0.4.6,<0.5
-emcee>=2.1.0
-george
-cython
-psutil
-pyDOE
-statsmodels
 joblib
-sobol_seq
-'''
+"""
+dependencies.verify_packages(__MANDATORY_PACKAGES__)
 
+if sys.version_info < (3, 5, 2):
+    raise ValueError("SMAC requires Python 3.5.2 or newer.")
 
-dependencies.verify_packages(__MANDATORY_PACKAGES__)
 
 if os.name != 'posix':
     print(

smac/configspace/__init__.py

@@ -1,8 +1,10 @@
 from functools import partial
 
-from ConfigSpace import ConfigurationSpace, Configuration, CategoricalHyperparameter
+from ConfigSpace import ConfigurationSpace, Configuration, Constant,\
+     CategoricalHyperparameter, UniformFloatHyperparameter, \
+     UniformIntegerHyperparameter, InCondition
 from ConfigSpace.read_and_write import pcs, pcs_new, json
 from ConfigSpace.util import get_one_exchange_neighbourhood
 from smac.configspace.util import convert_configurations_to_array
 
-get_one_exchange_neighbourhood = partial(get_one_exchange_neighbourhood, stdev=0.05, num_neighbors=8)
+get_one_exchange_neighbourhood = partial(get_one_exchange_neighbourhood, stdev=0.05, num_neighbors=8)

smac/optimizer/acquisition.py

@@ -340,3 +340,106 @@ def calculate_log_ei():
                 "Expected Improvement is smaller than 0 for at least one sample.")
 
         return log_ei.reshape((-1, 1))
+
+
+class PI(AbstractAcquisitionFunction):
+    def __init__(self,
+                 model: AbstractEPM,
+                 par: float=0.0):
+
+        """Computes the probability of improvement for a given x over the best so far value as
+        acquisition value.
+
+        :math:`P(f_{t+1}(\mathbf{X})\geq f(\mathbf{X^+})) :=
+        \Phi(\frac{\mu(\mathbf{X}) - f(\mathbf{X^+})}{\sigma(\mathbf{X})})`,
+        with :math:`f(X^+)` as the incumbent and :math:`\Phi` the cdf of the standard normal
+
+        Parameters
+        ----------
+        model : AbstractEPM
+            A model that implements at least
+                 - predict_marginalized_over_instances(X)
+        par : float, default=0.0
+            Controls the balance between exploration and exploitation of the
+            acquisition function.
+        """
+        super(PI, self).__init__(model)
+        self.long_name = 'Probability of Improvement'
+        self.par = par
+        self.eta = None
+
+    def _compute(self, X: np.ndarray):
+        """Computes the PI value.
+
+        Parameters
+        ----------
+        X: np.ndarray(N, D)
+           Points to evaluate PI. N is the number of points and D the dimension for the points
+
+        Returns
+        -------
+        np.ndarray(N,1)
+            Expected Improvement of X
+        """
+        if self.eta is None:
+            raise ValueError('No current best specified. Call update('
+                             'eta=<float>) to inform the acquisition function '
+                             'about the current best value.')
+
+        if len(X.shape) == 1:
+            X = X[:, np.newaxis]
+        m, var_ = self.model.predict_marginalized_over_instances(X)
+        std = np.sqrt(var_)
+        return norm.cdf((self.eta - m - self.par) / std)
+
+
+class LCB(AbstractAcquisitionFunction):
+    def __init__(self,
+                 model: AbstractEPM,
+                 par: float=1.0):
+
+        """Computes the lower confidence bound for a given x over the best so far value as
+        acquisition value.
+
+        :math:`LCB(X) = \mu(\mathbf{X}) - \sqrt(\beta_t)\sigma(\mathbf{X})`
+
+        Returns -LCB(X) as the acquisition_function optimizer maximizes the acquisition value.
+
+        Parameters
+        ----------
+        model : AbstractEPM
+            A model that implements at least
+                 - predict_marginalized_over_instances(X)
+        par : float, default=0.0
+            Controls the balance between exploration and exploitation of the
+            acquisition function.
+        """
+        super(LCB, self).__init__(model)
+        self.long_name = 'Lower Confidence Bound'
+        self.par = par
+        self.eta = None  # to be compatible with the existing update calls in SMBO
+        self.num_data = None
+
+    def _compute(self, X: np.ndarray):
+        """Computes the LCB value.
+
+        Parameters
+        ----------
+        X: np.ndarray(N, D)
+           Points to evaluate LCB. N is the number of points and D the dimension for the points
+
+        Returns
+        -------
+        np.ndarray(N,1)
+            Expected Improvement of X
+        """
+        if self.num_data is None:
+            raise ValueError('No current number of Datapoints specified. Call update('
+                             'num_data=<int>) to inform the acquisition function '
+                             'about the number of datapoints.')
+        if len(X.shape) == 1:
+            X = X[:, np.newaxis]
+        m, var_ = self.model.predict_marginalized_over_instances(X)
+        std = np.sqrt(var_)
+        beta = 2*np.log((X.shape[1] * self.num_data**2) / self.par)
+        return -(m - np.sqrt(beta)*std)