Permalink
650 lines (545 sloc) 25.6 KB
from __future__ import division
try:
import cPickle as pickle
except ImportError:
import pickle
from bisect import bisect_right
from collections import defaultdict
from copy import deepcopy
from functools import partial
from itertools import chain
from operator import eq
def identity(obj):
"""Returns directly the argument *obj*.
"""
return obj
class History(object):
"""The :class:`History` class helps to build a genealogy of all the
individuals produced in the evolution. It contains two attributes,
the :attr:`genealogy_tree` that is a dictionary of lists indexed by
individual, the list contain the indices of the parents. The second
attribute :attr:`genealogy_history` contains every individual indexed
by their individual number as in the genealogy tree.
The produced genealogy tree is compatible with `NetworkX
<http://networkx.lanl.gov/index.html>`_, here is how to plot the genealogy
tree ::
history = History()
# Decorate the variation operators
toolbox.decorate("mate", history.decorator)
toolbox.decorate("mutate", history.decorator)
# Create the population and populate the history
population = toolbox.population(n=POPSIZE)
history.update(population)
# Do the evolution, the decorators will take care of updating the
# history
# [...]
import matplotlib.pyplot as plt
import networkx
graph = networkx.DiGraph(history.genealogy_tree)
graph = graph.reverse() # Make the grah top-down
colors = [toolbox.evaluate(history.genealogy_history[i])[0] for i in graph]
networkx.draw(graph, node_color=colors)
plt.show()
Using NetworkX in combination with `pygraphviz
<http://networkx.lanl.gov/pygraphviz/>`_ (dot layout) this amazing
genealogy tree can be obtained from the OneMax example with a population
size of 20 and 5 generations, where the color of the nodes indicate there
fitness, blue is low and red is high.
.. image:: /_images/genealogy.png
:width: 67%
.. note::
The genealogy tree might get very big if your population and/or the
number of generation is large.
"""
def __init__(self):
self.genealogy_index = 0
self.genealogy_history = dict()
self.genealogy_tree = dict()
def update(self, individuals):
"""Update the history with the new *individuals*. The index present in
their :attr:`history_index` attribute will be used to locate their
parents, it is then modified to a unique one to keep track of those
new individuals. This method should be called on the individuals after
each variation.
:param individuals: The list of modified individuals that shall be
inserted in the history.
If the *individuals* do not have a :attr:`history_index` attribute,
the attribute is added and this individual is considered as having no
parent. This method should be called with the initial population to
initialize the history.
Modifying the internal :attr:`genealogy_index` of the history or the
:attr:`history_index` of an individual may lead to unpredictable
results and corruption of the history.
"""
try:
parent_indices = tuple(ind.history_index for ind in individuals)
except AttributeError:
parent_indices = tuple()
for ind in individuals:
self.genealogy_index += 1
ind.history_index = self.genealogy_index
self.genealogy_history[self.genealogy_index] = deepcopy(ind)
self.genealogy_tree[self.genealogy_index] = parent_indices
@property
def decorator(self):
"""Property that returns an appropriate decorator to enhance the
operators of the toolbox. The returned decorator assumes that the
individuals are returned by the operator. First the decorator calls
the underlying operation and then calls the :func:`update` function
with what has been returned by the operator. Finally, it returns the
individuals with their history parameters modified according to the
update function.
"""
def decFunc(func):
def wrapFunc(*args, **kargs):
individuals = func(*args, **kargs)
self.update(individuals)
return individuals
return wrapFunc
return decFunc
def getGenealogy(self, individual, max_depth=float("inf")):
"""Provide the genealogy tree of an *individual*. The individual must
have an attribute :attr:`history_index` as defined by
:func:`~deap.tools.History.update` in order to retrieve its associated
genealogy tree. The returned graph contains the parents up to
*max_depth* variations before this individual. If not provided
the maximum depth is up to the begining of the evolution.
:param individual: The individual at the root of the genealogy tree.
:param max_depth: The approximate maximum distance between the root
(individual) and the leaves (parents), optional.
:returns: A dictionary where each key is an individual index and the
values are a tuple corresponding to the index of the parents.
"""
gtree = {}
visited = set() # Adds memory to the breadth first search
def genealogy(index, depth):
if index not in self.genealogy_tree:
return
depth += 1
if depth > max_depth:
return
parent_indices = self.genealogy_tree[index]
gtree[index] = parent_indices
for ind in parent_indices:
if ind not in visited:
genealogy(ind, depth)
visited.add(ind)
genealogy(individual.history_index, 0)
return gtree
class Statistics(object):
"""Object that compiles statistics on a list of arbitrary objects.
When created the statistics object receives a *key* argument that
is used to get the values on which the function will be computed.
If not provided the *key* argument defaults to the identity function.
The value returned by the key may be a multi-dimensional object, i.e.:
a tuple or a list, as long as the statistical function registered
support it. So for example, statistics can be computed directly on
multi-objective fitnesses when using numpy statistical function.
:param key: A function to access the values on which to compute the
statistics, optional.
::
>>> s = Statistics()
>>> s.register("mean", numpy.mean)
>>> s.register("max", max)
>>> s.compile([1, 2, 3, 4])
{'max': 4, 'mean': 2.5}
>>> s.compile([5, 6, 7, 8])
{'max': 8, 'mean': 6.5}
"""
def __init__(self, key=identity):
self.key = key
self.functions = dict()
self.fields = []
def register(self, name, function, *args, **kargs):
"""Register a *function* that will be applied on the sequence each
time :meth:`record` is called.
:param name: The name of the statistics function as it would appear
in the dictionnary of the statistics object.
:param function: A function that will compute the desired statistics
on the data as preprocessed by the key.
:param argument: One or more argument (and keyword argument) to pass
automatically to the registered function when called,
optional.
"""
self.functions[name] = partial(function, *args, **kargs)
self.fields.append(name)
def compile(self, data):
"""Apply to the input sequence *data* each registered function
and return the results as a dictionnary.
:param data: Sequence of objects on which the statistics are computed.
"""
values = tuple(self.key(elem) for elem in data)
entry = dict()
for key, func in self.functions.iteritems():
entry[key] = func(values)
return entry
class MultiStatistics(dict):
"""Dictionary of :class:`Statistics` object allowing to compute
statistics on multiple keys using a single call to :meth:`compile`. It
takes a set of key-value pairs associating a statistics object to a
unique name. This name can then be used to retrieve the statistics object.
The following code computes statistics simultaneously on the length and
the first value of the provided objects.
::
>>> len_stats = Statistics(key=len)
>>> itm0_stats = Statistics(key=itemgetter(0))
>>> mstats = MultiStatistics(length=len_stats, item=itm0_stats)
>>> mstats.register("mean", numpy.mean, axis=0)
>>> mstats.register("max", numpy.max, axis=0)
>>> mstats.compile([[0.0, 1.0, 1.0, 5.0], [2.0, 5.0]])
{'length': {'max': 4, 'mean': 3.0}, 'item': {'max': 2.0, 'mean': 1.0}}
"""
def compile(self, data):
"""Calls :meth:`Statistics.compile` with *data* of each
:class:`Statistics` object.
:param data: Sequence of objects on which the statistics are computed.
"""
record = {}
for name, stats in self.items():
record[name] = stats.compile(data)
return record
@property
def fields(self):
return sorted(self.keys())
def register(self, name, function, *args, **kargs):
"""Register a *function* in each :class:`Statistics` object.
:param name: The name of the statistics function as it would appear
in the dictionnary of the statistics object.
:param function: A function that will compute the desired statistics
on the data as preprocessed by the key.
:param argument: One or more argument (and keyword argument) to pass
automatically to the registered function when called,
optional.
"""
for stats in self.values():
stats.register(name, function, *args, **kargs)
class Logbook(list):
"""Evolution records as a chronological list of dictionaries.
Data can be retrieved via the :meth:`select` method given the appropriate
names.
The :class:`Logbook` class may also contain other logbooks refered to
as chapters. Chapters are used to store information associated to a
specific part of the evolution. For example when computing statistics
on different components of individuals (namely :class:`MultiStatistics`),
chapters can be used to distinguish the average fitness and the average
size.
"""
def __init__(self):
self.buffindex = 0
self.chapters = defaultdict(Logbook)
"""Dictionary containing the sub-sections of the logbook which are also
:class:`Logbook`. Chapters are automatically created when the right hand
side of a keyworded argument, provided to the *record* function, is a
dictionnary. The keyword determines the chapter's name. For example, the
following line adds a new chapter "size" that will contain the fields
"max" and "mean". ::
logbook.record(gen=0, size={'max' : 10.0, 'mean' : 7.5})
To access a specific chapter, use the name of the chapter as a
dictionnary key. For example, to access the size chapter and select
the mean use ::
logbook.chapters["size"].select("mean")
Compiling a :class:`MultiStatistics` object returns a dictionary
containing dictionnaries, therefore when recording such an object in a
logbook using the keyword argument unpacking operator (**), chapters
will be automatically added to the logbook.
::
>>> fit_stats = Statistics(key=attrgetter("fitness.values"))
>>> size_stats = Statistics(key=len)
>>> mstats = MultiStatistics(fitness=fit_stats, size=size_stats)
>>> # [...]
>>> record = mstats.compile(population)
>>> logbook.record(**record)
>>> print logbook
fitness length
------------ ------------
max mean max mean
2 1 4 3
"""
self.columns_len = None
self.header = None
"""Order of the columns to print when using the :data:`stream` and
:meth:`__str__` methods. The syntax is a single iterable containing
string elements. For example, with the previously
defined statistics class, one can print the generation and the
fitness average, and maximum with
::
logbook.header = ("gen", "mean", "max")
If not set the header is built with all fields, in arbritrary order
on insertion of the first data. The header can be removed by setting
it to :data:`None`.
"""
self.log_header = True
"""Tells the log book to output or not the header when streaming the
first line or getting its entire string representation. This defaults
:data:`True`.
"""
def record(self, **infos):
"""Enter a record of event in the logbook as a list of key-value pairs.
The informations are appended chronogically to a list as a dictionnary.
When the value part of a pair is a dictionnary, the informations contained
in the dictionnary are recorded in a chapter entitled as the name of the
key part of the pair. Chapters are also Logbook.
"""
for key, value in infos.items():
if isinstance(value, dict):
self.chapters[key].record(**value)
del infos[key]
self.append(infos)
def select(self, *names):
"""Return a list of values associated to the *names* provided
in argument in each dictionary of the Statistics object list.
One list per name is returned in order.
::
>>> log = Logbook()
>>> log.record(gen = 0, mean = 5.4, max = 10.0)
>>> log.record(gen = 1, mean = 9.4, max = 15.0)
>>> log.select("mean")
[5.4, 9.4]
>>> log.select("gen", "max")
([0, 1], [10.0, 15.0])
With a :class:`MultiStatistics` object, the statistics for each
measurement can be retrieved using the :data:`chapters` member :
::
>>> log = Logbook()
>>> log.record(**{'gen' : 0, 'fit' : {'mean' : 0.8, 'max' : 1.5},
... 'size' : {'mean' : 25.4, 'max' : 67}})
>>> log.record(**{'gen' : 1, 'fit' : {'mean' : 0.95, 'max' : 1.7},
... 'size' : {'mean' : 28.1, 'max' : 71}})
>>> log.chapters['size'].select("mean")
[25.4, 28.1]
>>> log.chapters['fit'].select("gen", "max")
([0, 1], [1.5, 1.7])
"""
if len(names) == 1:
return [entry.get(names[0], None) for entry in self]
return tuple([entry.get(name, None) for entry in self] for name in names)
@property
def stream(self):
"""Retrieve the formatted not streamed yet entries of the database
including the headers.
::
>>> log = Logbook()
>>> log.append({'gen' : 0})
>>> print log.stream
gen
0
>>> log.append({'gen' : 1})
>>> print log.stream
1
"""
startindex, self.buffindex = self.buffindex, len(self)
return self.__str__(startindex)
def __delitem__(self, key):
if isinstance(key, slice):
for i, in range(*key.indices(len(self))):
self.pop(i)
for chapter in self.chapters.values():
chapter.pop(i)
else:
self.pop(key)
for chapter in self.chapters.values():
chapter.pop(key)
def pop(self, index=0):
"""Retrieve and delete element *index*. The header and stream will be
adjusted to follow the modification.
:param item: The index of the element to remove, optional. It defaults
to the first element.
You can also use the following syntax to delete elements.
::
del log[0]
del log[1::5]
"""
if index < self.buffindex:
self.buffindex -= 1
return super(self.__class__, self).pop(index)
def __txt__(self, startindex):
columns = self.header
if not columns:
columns = sorted(self[0].keys()) + sorted(self.chapters.keys())
if not self.columns_len or len(self.columns_len) != len(columns):
self.columns_len = map(len, columns)
chapters_txt = {}
offsets = defaultdict(int)
for name, chapter in self.chapters.items():
chapters_txt[name] = chapter.__txt__(startindex)
if startindex == 0:
offsets[name] = len(chapters_txt[name]) - len(self)
str_matrix = []
for i, line in enumerate(self[startindex:]):
str_line = []
for j, name in enumerate(columns):
if name in chapters_txt:
column = chapters_txt[name][i+offsets[name]]
else:
value = line.get(name, "")
string = "{0:n}" if isinstance(value, float) else "{0}"
column = string.format(value)
self.columns_len[j] = max(self.columns_len[j], len(column))
str_line.append(column)
str_matrix.append(str_line)
if startindex == 0 and self.log_header:
header = []
nlines = 1
if len(self.chapters) > 0:
nlines += max(map(len, chapters_txt.values())) - len(self) + 1
header = [[] for i in xrange(nlines)]
for j, name in enumerate(columns):
if name in chapters_txt:
length = max(len(line.expandtabs()) for line in chapters_txt[name])
blanks = nlines - 2 - offsets[name]
for i in xrange(blanks):
header[i].append(" " * length)
header[blanks].append(name.center(length))
header[blanks+1].append("-" * length)
for i in xrange(offsets[name]):
header[blanks+2+i].append(chapters_txt[name][i])
else:
length = max(len(line[j].expandtabs()) for line in str_matrix)
for line in header[:-1]:
line.append(" " * length)
header[-1].append(name)
str_matrix = chain(header, str_matrix)
template = "\t".join("{%i:<%i}" % (i, l) for i, l in enumerate(self.columns_len))
text = [template.format(*line) for line in str_matrix]
return text
def __str__(self, startindex=0):
text = self.__txt__(startindex)
return "\n".join(text)
class HallOfFame(object):
"""The hall of fame contains the best individual that ever lived in the
population during the evolution. It is lexicographically sorted at all
time so that the first element of the hall of fame is the individual that
has the best first fitness value ever seen, according to the weights
provided to the fitness at creation time.
The insertion is made so that old individuals have priority on new
individuals. A single copy of each individual is kept at all time, the
equivalence between two individuals is made by the operator passed to the
*similar* argument.
:param maxsize: The maximum number of individual to keep in the hall of
fame.
:param similar: An equivalence operator between two individuals, optional.
It defaults to operator :func:`operator.eq`.
The class :class:`HallOfFame` provides an interface similar to a list
(without being one completely). It is possible to retrieve its length, to
iterate on it forward and backward and to get an item or a slice from it.
"""
def __init__(self, maxsize, similar=eq):
self.maxsize = maxsize
self.keys = list()
self.items = list()
self.similar = similar
def update(self, population):
"""Update the hall of fame with the *population* by replacing the
worst individuals in it by the best individuals present in
*population* (if they are better). The size of the hall of fame is
kept constant.
:param population: A list of individual with a fitness attribute to
update the hall of fame with.
"""
if len(self) == 0 and self.maxsize !=0 and len(population) > 0:
# Working on an empty hall of fame is problematic for the
# "for else"
self.insert(population[0])
for ind in population:
if ind.fitness > self[-1].fitness or len(self) < self.maxsize:
for hofer in self:
# Loop through the hall of fame to check for any
# similar individual
if self.similar(ind, hofer):
break
else:
# The individual is unique and strictly better than
# the worst
if len(self) >= self.maxsize:
self.remove(-1)
self.insert(ind)
def insert(self, item):
"""Insert a new individual in the hall of fame using the
:func:`~bisect.bisect_right` function. The inserted individual is
inserted on the right side of an equal individual. Inserting a new
individual in the hall of fame also preserve the hall of fame's order.
This method **does not** check for the size of the hall of fame, in a
way that inserting a new individual in a full hall of fame will not
remove the worst individual to maintain a constant size.
:param item: The individual with a fitness attribute to insert in the
hall of fame.
"""
item = deepcopy(item)
i = bisect_right(self.keys, item.fitness)
self.items.insert(len(self) - i, item)
self.keys.insert(i, item.fitness)
def remove(self, index):
"""Remove the specified *index* from the hall of fame.
:param index: An integer giving which item to remove.
"""
del self.keys[len(self) - (index % len(self) + 1)]
del self.items[index]
def clear(self):
"""Clear the hall of fame."""
del self.items[:]
del self.keys[:]
def __len__(self):
return len(self.items)
def __getitem__(self, i):
return self.items[i]
def __iter__(self):
return iter(self.items)
def __reversed__(self):
return reversed(self.items)
def __str__(self):
return str(self.items)
class ParetoFront(HallOfFame):
"""The Pareto front hall of fame contains all the non-dominated individuals
that ever lived in the population. That means that the Pareto front hall of
fame can contain an infinity of different individuals.
:param similar: A function that tels the Pareto front whether or not two
individuals are similar, optional.
The size of the front may become very large if it is used for example on
a continuous function with a continuous domain. In order to limit the number
of individuals, it is possible to specify a similarity function that will
return :data:`True` if the genotype of two individuals are similar. In that
case only one of the two individuals will be added to the hall of fame. By
default the similarity function is :func:`operator.eq`.
Since, the Pareto front hall of fame inherits from the :class:`HallOfFame`,
it is sorted lexicographically at every moment.
"""
def __init__(self, similar=eq):
HallOfFame.__init__(self, None, similar)
def update(self, population):
"""Update the Pareto front hall of fame with the *population* by adding
the individuals from the population that are not dominated by the hall
of fame. If any individual in the hall of fame is dominated it is
removed.
:param population: A list of individual with a fitness attribute to
update the hall of fame with.
"""
for ind in population:
is_dominated = False
dominates_one = False
has_twin = False
to_remove = []
for i, hofer in enumerate(self): # hofer = hall of famer
if not dominates_one and hofer.fitness.dominates(ind.fitness):
is_dominated = True
break
elif ind.fitness.dominates(hofer.fitness):
dominates_one = True
to_remove.append(i)
elif ind.fitness == hofer.fitness and self.similar(ind, hofer):
has_twin = True
break
for i in reversed(to_remove): # Remove the dominated hofer
self.remove(i)
if not is_dominated and not has_twin:
self.insert(ind)
__all__ = ['HallOfFame', 'ParetoFront', 'History', 'Statistics', 'MultiStatistics', 'Logbook']
if __name__ == "__main__":
import doctest
from operator import itemgetter
import numpy
doctest.run_docstring_examples(Statistics, globals())
doctest.run_docstring_examples(Statistics.register, globals())
doctest.run_docstring_examples(Statistics.compile, globals())
doctest.run_docstring_examples(MultiStatistics, globals())
doctest.run_docstring_examples(MultiStatistics.register, globals())
doctest.run_docstring_examples(MultiStatistics.compile, globals())