#! /usr/bin/env python
# Copyright (c) 2014 KU Leuven, ESAT-STADIUS
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions
# are met:
#
# 1. Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
#
# 2. Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
#
# 3. Neither name of copyright holders nor the names of its contributors
# may be used to endorse or promote products derived from this software
# without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
# ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
"""A collection of top-level API functions for Optunity.
Main functions in this module:
* :func:`make_solver`
* :func:`suggest_solver`
* :func:`manual`
* :func:`maximize`
* :func:`maximize_structured`
* :func:`minimize`
* :func:`minimize_structured`
* :func:`optimize`
We recommend using these functions rather than equivalents found in other places,
e.g. :mod:`optunity.solvers`.
.. moduleauthor:: Marc Claesen
"""
import timeit
import sys
import operator
# optunity imports
from . import functions as fun
from . import solvers
from . import search_spaces
from .solvers import solver_registry
from .util import DocumentedNamedTuple as DocTup
from .constraints import wrap_constraints
def _manual_lines(solver_name=None):
"""Brief solver manual.
:param solver_name: (optional) name of the solver to request a manual from.
If none is specified, a general manual and list of all registered solvers is returned.
:result:
* list of strings that contain the requested manual
* solver name(s): name of the solver that was specified or list of all registered solvers.
Raises ``KeyError`` if ``solver_name`` is not registered."""
if solver_name:
return solver_registry.get(solver_name).desc_full, [solver_name]
else:
return solver_registry.manual(), solver_registry.solver_names()
[docs]def available_solvers():
"""Returns a list of all available solvers.
These can be used in :func:`optunity.make_solver`.
"""
return solver_registry.solver_names()
[docs]def manual(solver_name=None):
"""Prints the manual of requested solver.
:param solver_name: (optional) name of the solver to request a manual from.
If none is specified, a general manual is printed.
Raises ``KeyError`` if ``solver_name`` is not registered."""
if solver_name:
man = solver_registry.get(solver_name).desc_full
else:
man = solver_registry.manual()
print('\n'.join(man))
optimize_results = DocTup("""
**Result details includes the following**:
optimum
optimal function value f(solution)
stats
statistics about the solving process
call_log
the call log
report
solver report, can be None
""",
'optimize_results', ['optimum',
'stats',
'call_log',
'report']
)
optimize_stats = DocTup("""
**Statistics gathered while solving a problem**:
num_evals
number of function evaluations
time
wall clock time needed to solve
""",
'optimize_stats', ['num_evals', 'time'])
[docs]def suggest_solver(num_evals=50, solver_name=None, **kwargs):
if solver_name:
solvercls = solver_registry.get(solver_name)
else:
solver_name = 'particle swarm'
solvercls = solvers.ParticleSwarm
if hasattr(solvercls, 'suggest_from_box'):
suggestion = solvercls.suggest_from_box(num_evals, **kwargs)
elif hasattr(solvercls, 'suggest_from_seed'):
# the seed will be the center of the box that is provided to us
seed = dict([(k, float(v[0] + v[1]) / 2) for k, v in kwargs.items()])
suggestion = solvercls.suggest_from_seed(num_evals, **seed)
else:
raise ValueError('Unable to instantiate ' + solvercls.name + '.')
suggestion['solver_name'] = solver_name
return suggestion
[docs]def maximize(f, num_evals=50, solver_name=None, pmap=map, **kwargs):
"""Basic function maximization routine. Maximizes ``f`` within
the given box constraints.
:param f: the function to be maximized
:param num_evals: number of permitted function evaluations
:param solver_name: name of the solver to use (optional)
:type solver_name: string
:param pmap: the map function to use
:type pmap: callable
:param kwargs: box constraints, a dict of the following form
``{'parameter_name': [lower_bound, upper_bound], ...}``
:returns: retrieved maximum, extra information and solver info
This function will implicitly choose an appropriate solver and
its initialization based on ``num_evals`` and the box constraints.
"""
# sanity check on box constraints
assert all([len(v) == 2 and v[0] < v[1]
for v in kwargs.values()]), 'Box constraints improperly specified: should be [lb, ub] pairs'
f = _wrap_hard_box_constraints(f, kwargs, -sys.float_info.max)
suggestion = suggest_solver(num_evals, solver_name, **kwargs)
solver = make_solver(**suggestion)
solution, details = optimize(solver, f, maximize=True, max_evals=num_evals,
pmap=pmap)
return solution, details, suggestion
[docs]def minimize(f, num_evals=50, solver_name=None, pmap=map, **kwargs):
"""Basic function minimization routine. Minimizes ``f`` within
the given box constraints.
:param f: the function to be minimized
:param num_evals: number of permitted function evaluations
:param solver_name: name of the solver to use (optional)
:type solver_name: string
:param pmap: the map function to use
:type pmap: callable
:param kwargs: box constraints, a dict of the following form
``{'parameter_name': [lower_bound, upper_bound], ...}``
:returns: retrieved minimum, extra information and solver info
This function will implicitly choose an appropriate solver and
its initialization based on ``num_evals`` and the box constraints.
"""
# sanity check on box constraints
assert all([len(v) == 2 and v[0] < v[1]
for v in kwargs.values()]), 'Box constraints improperly specified: should be [lb, ub] pairs'
func = _wrap_hard_box_constraints(f, kwargs, sys.float_info.max)
suggestion = suggest_solver(num_evals, solver_name, **kwargs)
solver = make_solver(**suggestion)
solution, details = optimize(solver, func, maximize=False, max_evals=num_evals,
pmap=pmap)
return solution, details, suggestion
[docs]def optimize(solver, func, maximize=True, max_evals=0, pmap=map, decoder=None):
"""Optimizes func with given solver.
:param solver: the solver to be used, for instance a result from :func:`optunity.make_solver`
:param func: the objective function
:type func: callable
:param maximize: maximize or minimize?
:type maximize: bool
:param max_evals: maximum number of permitted function evaluations
:type max_evals: int
:param pmap: the map() function to use, to vectorize use :func:`optunity.parallel.pmap`
:type pmap: function
Returns the solution and a namedtuple with further details.
Please refer to docs of optunity.maximize_results
and optunity.maximize_stats.
"""
if max_evals > 0:
f = fun.max_evals(max_evals)(func)
else:
f = func
f = fun.logged(f)
num_evals = -len(f.call_log)
time = timeit.default_timer()
try:
solution, report = solver.optimize(f, maximize, pmap=pmap)
except fun.MaximumEvaluationsException:
# early stopping because maximum number of evaluations is reached
# retrieve solution from the call log
report = None
if maximize:
index, _ = max(enumerate(f.call_log.values()), key=operator.itemgetter(1))
else:
index, _ = min(enumerate(f.call_log.values()), key=operator.itemgetter(1))
solution = list(f.call_log.keys())[index]._asdict()
time = timeit.default_timer() - time
# TODO why is this necessary?
if decoder: solution = decoder(solution)
optimum = f.call_log.get(**solution)
num_evals += len(f.call_log)
# use namedtuple to enforce uniformity in case of changes
stats = optimize_stats(num_evals, time)
call_dict = f.call_log.to_dict()
return solution, optimize_results(optimum, stats._asdict(),
call_dict, report)
optimize.__doc__ = '''
Optimizes func with given solver.
:param solver: the solver to be used, for instance a result from :func:`optunity.make_solver`
:param func: the objective function
:type func: callable
:param maximize: maximize or minimize?
:type maximize: bool
:param max_evals: maximum number of permitted function evaluations
:type max_evals: int
:param pmap: the map() function to use, to vectorize use :func:`optunity.pmap`
:type pmap: function
Returns the solution and a ``namedtuple`` with further details.
''' + optimize_results.__doc__ + optimize_stats.__doc__
[docs]def make_solver(solver_name, *args, **kwargs):
"""Creates a Solver from given parameters.
:param solver_name: the solver to instantiate
:type solver_name: string
:param args: positional arguments to solver constructor.
:param kwargs: keyword arguments to solver constructor.
Use :func:`optunity.manual` to get a list of registered solvers.
For constructor arguments per solver, please refer to :doc:`/user/solvers`.
Raises ``KeyError`` if
- ``solver_name`` is not registered
- ``*args`` and ``**kwargs`` are invalid to instantiate the solver.
"""
solvercls = solver_registry.get(solver_name)
return solvercls(*args, **kwargs)
[docs]def wrap_call_log(f, call_dict):
"""Wraps an existing call log (as dictionary) around f.
This allows you to communicate known function values to solvers.
(currently available solvers do not use this info)
"""
f = fun.logged(f)
call_log = fun.CallLog.from_dict(call_dict)
if f.call_log:
f.call_log.update(call_log)
else:
f.call_log = call_log
return f
def _wrap_hard_box_constraints(f, box, default):
"""Places hard box constraints on the domain of ``f``
and defaults function values if constraints are violated.
:param f: the function to be wrapped with constraints
:type f: callable
:param box: the box, as a dict: ``{'param_name': [lb, ub], ...}``
:type box: dict
:param default: function value to default to when constraints
are violated
:type default: number
"""
return wrap_constraints(f, default, range_oo=box)
[docs]def maximize_structured(f, search_space, num_evals=50, pmap=map):
"""Basic function maximization routine. Maximizes ``f`` within
the given box constraints.
:param f: the function to be maximized
:param search_space: the search space (see :doc:`/user/structured_search_spaces` for details)
:param num_evals: number of permitted function evaluations
:param pmap: the map function to use
:type pmap: callable
:returns: retrieved maximum, extra information and solver info
This function will implicitly choose an appropriate solver and
its initialization based on ``num_evals`` and the box constraints.
"""
tree = search_spaces.SearchTree(search_space)
box = tree.to_box()
# we need to position the call log here
# because the function signature used later on is internal logic
f = fun.logged(f)
# wrap the decoder and constraints for the internal search space representation
f = tree.wrap_decoder(f)
f = _wrap_hard_box_constraints(f, box, -sys.float_info.max)
suggestion = suggest_solver(num_evals, "particle swarm", **box)
solver = make_solver(**suggestion)
solution, details = optimize(solver, f, maximize=True, max_evals=num_evals,
pmap=pmap, decoder=tree.decode)
return solution, details, suggestion
[docs]def minimize_structured(f, search_space, num_evals=50, pmap=map):
"""Basic function minimization routine. Minimizes ``f`` within
the given box constraints.
:param f: the function to be maximized
:param search_space: the search space (see :doc:`/user/structured_search_spaces` for details)
:param num_evals: number of permitted function evaluations
:param pmap: the map function to use
:type pmap: callable
:returns: retrieved maximum, extra information and solver info
This function will implicitly choose an appropriate solver and
its initialization based on ``num_evals`` and the box constraints.
"""
tree = search_spaces.SearchTree(search_space)
box = tree.to_box()
# we need to position the call log here
# because the function signature used later on is internal logic
f = fun.logged(f)
# wrap the decoder and constraints for the internal search space representation
f = tree.wrap_decoder(f)
f = _wrap_hard_box_constraints(f, box, sys.float_info.max)
suggestion = suggest_solver(num_evals, "particle swarm", **box)
solver = make_solver(**suggestion)
solution, details = optimize(solver, f, maximize=False, max_evals=num_evals,
pmap=pmap, decoder=tree.decode)
return solution, details, suggestion