wangwei
/
graphkit-learn
Not watched
1
0
Fork 0
Code
Releases 2 Wiki Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain
Browse Source

[Major Feature] Graph kernel classes now can compute kernel matrix between two different list of graphs using fit/transform methods which uses the same scheme as the scikit-learn library!

v0.2.x
jajupmochi 4 years ago
parent
d08eddc69c
commit
bb36d0f507
12 changed files with 825 additions and 254 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +5
    -5
      gklearn/kernels/common_walk.py
  2. +13
    -13
      gklearn/kernels/conjugate_gradient.py
  3. +13
    -13
      gklearn/kernels/fixed_point.py
  4. +379
    -34
      gklearn/kernels/graph_kernel.py
  5. +14
    -14
      gklearn/kernels/marginalized.py
  6. +14
    -14
      gklearn/kernels/path_up_to_h.py
  7. +14
    -14
      gklearn/kernels/shortest_path.py
  8. +19
    -19
      gklearn/kernels/spectral_decomposition.py
  9. +14
    -14
      gklearn/kernels/structural_sp.py
  10. +18
    -18
      gklearn/kernels/sylvester_equation.py
  11. +313
    -87
      gklearn/kernels/treelet.py
  12. +9
    -9
      gklearn/kernels/weisfeiler_lehman.py

+ 5
- 5
gklearn/kernels/common_walk.py View File

@@ -47,7 +47,7 @@ class CommonWalk(GraphKernel):
itr = combinations_with_replacement(range(0, len(self._graphs)), 2)
len_itr = int(len(self._graphs) * (len(self._graphs) + 1) / 2)
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout,
length=len_itr, verbose=(self._verbose >= 2))
length=len_itr, verbose=(self.verbose >= 2))

# direct product graph method - exponential
if self._compute_method == 'exp':
@@ -86,7 +86,7 @@ class CommonWalk(GraphKernel):
do_fun = self._wrapper_kernel_do_geo

parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=_init_worker_gm,
glbv=(self._graphs,), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(self._graphs,), n_jobs=self.n_jobs, verbose=self.verbose)

return gram_matrix

@@ -100,9 +100,9 @@ class CommonWalk(GraphKernel):

# compute kernel list.
kernel_list = [None] * len(g_list)
if self._verbose >= 2:
if self.verbose >= 2:
iterator = get_iters(range(len(g_list)), desc='Computing kernels',
file=sys.stdout, length=len(g_list), verbose=(self._verbose >= 2))
file=sys.stdout, length=len(g_list), verbose=(self.verbose >= 2))
else:
iterator = range(len(g_list))

@@ -148,7 +148,7 @@ class CommonWalk(GraphKernel):
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=_init_worker_list, glbv=(g1, g_list), method='imap_unordered',
n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)

return kernel_list



+ 13
- 13
gklearn/kernels/conjugate_gradient.py View File

@@ -35,7 +35,7 @@ class ConjugateGradient(RandomWalkMeta):


def _compute_gm_series(self):
self._check_edge_weight(self._graphs, self._verbose)
self._check_edge_weight(self._graphs, self.verbose)
self._check_graphs(self._graphs)

lmda = self._weight
@@ -44,7 +44,7 @@ class ConjugateGradient(RandomWalkMeta):
gram_matrix = np.zeros((len(self._graphs), len(self._graphs)))

# Reindex nodes using consecutive integers for the convenience of kernel computation.
iterator = get_iters(self._graphs, desc='Reindex vertices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='Reindex vertices', file=sys.stdout, verbose=(self.verbose >= 2))
self._graphs = [nx.convert_node_labels_to_integers(g, first_label=0, label_attribute='label_orignal') for g in iterator]

if self._p is None and self._q is None: # p and q are uniform distributions as default.
@@ -52,7 +52,7 @@ class ConjugateGradient(RandomWalkMeta):
from itertools import combinations_with_replacement
itr = combinations_with_replacement(range(0, len(self._graphs)), 2)
len_itr = int(len(self._graphs) * (len(self._graphs) + 1) / 2)
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout, length=len_itr, verbose=(self._verbose >= 2))
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout, length=len_itr, verbose=(self.verbose >= 2))

for i, j in iterator:
kernel = self._kernel_do(self._graphs[i], self._graphs[j], lmda)
@@ -66,7 +66,7 @@ class ConjugateGradient(RandomWalkMeta):


def _compute_gm_imap_unordered(self):
self._check_edge_weight(self._graphs, self._verbose)
self._check_edge_weight(self._graphs, self.verbose)
self._check_graphs(self._graphs)

# Compute Gram matrix.
@@ -74,7 +74,7 @@ class ConjugateGradient(RandomWalkMeta):

# @todo: parallel this.
# Reindex nodes using consecutive integers for the convenience of kernel computation.
iterator = get_iters(self._graphs, desc='Reindex vertices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='Reindex vertices', file=sys.stdout, verbose=(self.verbose >= 2))
self._graphs = [nx.convert_node_labels_to_integers(g, first_label=0, label_attribute='label_orignal') for g in iterator]

if self._p is None and self._q is None: # p and q are uniform distributions as default.
@@ -86,7 +86,7 @@ class ConjugateGradient(RandomWalkMeta):
do_fun = self._wrapper_kernel_do

parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=init_worker,
glbv=(self._graphs,), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(self._graphs,), n_jobs=self.n_jobs, verbose=self.verbose)

else: # @todo
pass
@@ -95,7 +95,7 @@ class ConjugateGradient(RandomWalkMeta):


def _compute_kernel_list_series(self, g1, g_list):
self._check_edge_weight(g_list + [g1], self._verbose)
self._check_edge_weight(g_list + [g1], self.verbose)
self._check_graphs(g_list + [g1])

lmda = self._weight
@@ -105,11 +105,11 @@ class ConjugateGradient(RandomWalkMeta):

# Reindex nodes using consecutive integers for the convenience of kernel computation.
g1 = nx.convert_node_labels_to_integers(g1, first_label=0, label_attribute='label_orignal')
iterator = get_iters(g_list, desc='Reindex vertices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(g_list, desc='Reindex vertices', file=sys.stdout, verbose=(self.verbose >= 2))
g_list = [nx.convert_node_labels_to_integers(g, first_label=0, label_attribute='label_orignal') for g in iterator]

if self._p is None and self._q is None: # p and q are uniform distributions as default.
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self._verbose >= 2))
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self.verbose >= 2))

for i in iterator:
kernel = self._kernel_do(g1, g_list[i], lmda)
@@ -122,7 +122,7 @@ class ConjugateGradient(RandomWalkMeta):


def _compute_kernel_list_imap_unordered(self, g1, g_list):
self._check_edge_weight(g_list + [g1], self._verbose)
self._check_edge_weight(g_list + [g1], self.verbose)
self._check_graphs(g_list + [g1])

# compute kernel list.
@@ -131,7 +131,7 @@ class ConjugateGradient(RandomWalkMeta):
# Reindex nodes using consecutive integers for the convenience of kernel computation.
g1 = nx.convert_node_labels_to_integers(g1, first_label=0, label_attribute='label_orignal')
# @todo: parallel this.
iterator = get_iters(g_list, desc='Reindex vertices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(g_list, desc='Reindex vertices', file=sys.stdout, verbose=(self.verbose >= 2))
g_list = [nx.convert_node_labels_to_integers(g, first_label=0, label_attribute='label_orignal') for g in iterator]

if self._p is None and self._q is None: # p and q are uniform distributions as default.
@@ -149,7 +149,7 @@ class ConjugateGradient(RandomWalkMeta):
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=init_worker, glbv=(g1, g_list), method='imap_unordered',
n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)

else: # @todo
pass
@@ -162,7 +162,7 @@ class ConjugateGradient(RandomWalkMeta):


def _compute_single_kernel_series(self, g1, g2):
self._check_edge_weight([g1] + [g2], self._verbose)
self._check_edge_weight([g1] + [g2], self.verbose)
self._check_graphs([g1] + [g2])

lmda = self._weight


+ 13
- 13
gklearn/kernels/fixed_point.py View File

@@ -35,7 +35,7 @@ class FixedPoint(RandomWalkMeta):


def _compute_gm_series(self):
self._check_edge_weight(self._graphs, self._verbose)
self._check_edge_weight(self._graphs, self.verbose)
self._check_graphs(self._graphs)

lmda = self._weight
@@ -44,7 +44,7 @@ class FixedPoint(RandomWalkMeta):
gram_matrix = np.zeros((len(self._graphs), len(self._graphs)))

# Reindex nodes using consecutive integers for the convenience of kernel computation.
iterator = get_iters(self._graphs, desc='Reindex vertices', file=sys.stdout,verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='Reindex vertices', file=sys.stdout,verbose=(self.verbose >= 2))
self._graphs = [nx.convert_node_labels_to_integers(g, first_label=0, label_attribute='label_orignal') for g in iterator]

if self._p is None and self._q is None: # p and q are uniform distributions as default.
@@ -52,7 +52,7 @@ class FixedPoint(RandomWalkMeta):
from itertools import combinations_with_replacement
itr = combinations_with_replacement(range(0, len(self._graphs)), 2)
len_itr = int(len(self._graphs) * (len(self._graphs) + 1) / 2)
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout, length=len_itr, verbose=(self._verbose >= 2))
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout, length=len_itr, verbose=(self.verbose >= 2))

for i, j in iterator:
kernel = self._kernel_do(self._graphs[i], self._graphs[j], lmda)
@@ -66,7 +66,7 @@ class FixedPoint(RandomWalkMeta):


def _compute_gm_imap_unordered(self):
self._check_edge_weight(self._graphs, self._verbose)
self._check_edge_weight(self._graphs, self.verbose)
self._check_graphs(self._graphs)

# Compute Gram matrix.
@@ -74,7 +74,7 @@ class FixedPoint(RandomWalkMeta):

# @todo: parallel this.
# Reindex nodes using consecutive integers for the convenience of kernel computation.
iterator = get_iters(self._graphs, desc='Reindex vertices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='Reindex vertices', file=sys.stdout, verbose=(self.verbose >= 2))
self._graphs = [nx.convert_node_labels_to_integers(g, first_label=0, label_attribute='label_orignal') for g in iterator]

if self._p is None and self._q is None: # p and q are uniform distributions as default.
@@ -86,7 +86,7 @@ class FixedPoint(RandomWalkMeta):
do_fun = self._wrapper_kernel_do

parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=init_worker,
glbv=(self._graphs,), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(self._graphs,), n_jobs=self.n_jobs, verbose=self.verbose)

else: # @todo
pass
@@ -95,7 +95,7 @@ class FixedPoint(RandomWalkMeta):


def _compute_kernel_list_series(self, g1, g_list):
self._check_edge_weight(g_list + [g1], self._verbose)
self._check_edge_weight(g_list + [g1], self.verbose)
self._check_graphs(g_list + [g1])

lmda = self._weight
@@ -105,12 +105,12 @@ class FixedPoint(RandomWalkMeta):

# Reindex nodes using consecutive integers for the convenience of kernel computation.
g1 = nx.convert_node_labels_to_integers(g1, first_label=0, label_attribute='label_orignal')
iterator = get_iters(g_list, desc='Reindex vertices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(g_list, desc='Reindex vertices', file=sys.stdout, verbose=(self.verbose >= 2))
g_list = [nx.convert_node_labels_to_integers(g, first_label=0, label_attribute='label_orignal') for g in iterator]

if self._p is None and self._q is None: # p and q are uniform distributions as default.

iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self._verbose >= 2))
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self.verbose >= 2))

for i in iterator:
kernel = self._kernel_do(g1, g_list[i], lmda)
@@ -123,7 +123,7 @@ class FixedPoint(RandomWalkMeta):


def _compute_kernel_list_imap_unordered(self, g1, g_list):
self._check_edge_weight(g_list + [g1], self._verbose)
self._check_edge_weight(g_list + [g1], self.verbose)
self._check_graphs(g_list + [g1])

# compute kernel list.
@@ -132,7 +132,7 @@ class FixedPoint(RandomWalkMeta):
# Reindex nodes using consecutive integers for the convenience of kernel computation.
g1 = nx.convert_node_labels_to_integers(g1, first_label=0, label_attribute='label_orignal')
# @todo: parallel this.
iterator = get_iters(g_list, desc='Reindex vertices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(g_list, desc='Reindex vertices', file=sys.stdout, verbose=(self.verbose >= 2))
g_list = [nx.convert_node_labels_to_integers(g, first_label=0, label_attribute='label_orignal') for g in iterator]

if self._p is None and self._q is None: # p and q are uniform distributions as default.
@@ -150,7 +150,7 @@ class FixedPoint(RandomWalkMeta):
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=init_worker, glbv=(g1, g_list), method='imap_unordered',
n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)

else: # @todo
pass
@@ -163,7 +163,7 @@ class FixedPoint(RandomWalkMeta):


def _compute_single_kernel_series(self, g1, g2):
self._check_edge_weight([g1] + [g2], self._verbose)
self._check_edge_weight([g1] + [g2], self.verbose)
self._check_graphs([g1] + [g2])

lmda = self._weight


+ 379
- 34
gklearn/kernels/graph_kernel.py View File

@@ -9,27 +9,372 @@ import numpy as np
import networkx as nx
import multiprocessing
import time
# from abc import ABC, abstractmethod
from sklearn.base import BaseEstimator # , TransformerMixin
from sklearn.utils.validation import check_is_fitted # check_X_y, check_array,
from sklearn.exceptions import NotFittedError
from gklearn.utils import normalize_gram_matrix


class GraphKernel(object):
class GraphKernel(BaseEstimator): #, ABC):
"""The basic graph kernel class.

def __init__(self):
self._graphs = None
self._parallel = ''
self._n_jobs = 0
self._verbose = None
self._normalize = True
self._run_time = 0
self._gram_matrix = None
self._gram_matrix_unnorm = None
Attributes
----------
_graphs : list
Stores the input graphs on fit input data.
Default format of the list objects is `NetworkX` graphs.
**We don't guarantee that the input graphs remain unchanged during the
computation.**

References
----------
https://ysig.github.io/GraKeL/0.1a8/_modules/grakel/kernels/kernel.html#Kernel.
"""

def __init__(self, parallel=None, n_jobs=None, chunksize=None, normalize=True, verbose=2):
"""`__init__` for `GraphKernel` object."""
# @todo: the default settings of the parameters are different from those in the self.compute method.
# self._graphs = None
self.parallel = parallel
self.n_jobs = n_jobs
self.chunksize = chunksize
self.normalize = normalize
self.verbose = verbose
# self._run_time = 0
# self._gram_matrix = None
# self._gram_matrix_unnorm = None


##########################################################################
# The following is the 1st paradigm to compute kernel matrix, which is
# compatible with `scikit-learn`.
# -------------------------------------------------------------------
# Special thanks to the "GraKeL" library for providing an excellent template!
##########################################################################


def fit(self, X, y=None):
"""Fit a graph dataset for a transformer.

Parameters
----------
X : iterable
DESCRIPTION.

y : None, optional
There is no need of a target in a transformer, yet the `scikit-learn`
pipeline API requires this parameter.

Returns
-------
object
Returns self.

"""
# self._is_tranformed = False

# Clear any prior attributes stored on the estimator, # @todo: unless warm_start is used;
self.clear_attributes()

# X = check_array(X, accept_sparse=True)

# Validate parameters for the transformer.
self.validate_parameters()

# Validate the input.
self._graphs = self.validate_input(X)

# self._X = X
# self._kernel = self._get_kernel_instance()

# Return the transformer.
return self


def transform(self, X):
"""Compute the graph kernel matrix between given and fitted data.

Parameters
----------
X : TYPE
DESCRIPTION.

Raises
------
ValueError
DESCRIPTION.

Returns
-------
None.

"""
# Check if method "fit" had been called.
check_is_fitted(self, '_graphs')

# Validate the input.
Y = self.validate_input(X)

# Transform: compute the graph kernel matrix.
kernel_matrix = self.compute_kernel_matrix(Y)
self._Y = Y

# Self transform must appear before the diagonal call on normilization.
self._is_transformed = True
if self.normalize:
X_diag, Y_diag = self.diagonals()
kernel_matrix /= np.sqrt(np.outer(Y_diag, X_diag))

return kernel_matrix



def fit_transform(self, X):
"""Fit and transform: compute Gram matrix on the same data.

Parameters
----------
X : list of graphs
Input graphs.

Returns
-------
gram_matrix : numpy array, shape = [len(X), len(X)]
The Gram matrix of X.

"""
self.fit(X)

# Transform: compute Gram matrix.
gram_matrix = self.compute_kernel_matrix()

# Normalize.
self._X_diag = np.diagonal(gram_matrix).copy()
if self.normalize:
gram_matrix /= np.sqrt(np.outer(self._X_diag, self._X_diag))

return gram_matrix


def get_params(self):
pass


def set_params(self):
pass


def clear_attributes(self):
if hasattr(self, '_X_diag'):
delattr(self, '_X_diag')
if hasattr(self, '_graphs'):
delattr(self, '_graphs')
if hasattr(self, '_Y'):
delattr(self, '_Y')
if hasattr(self, '_run_time'):
delattr(self, '_run_time')


def validate_parameters(self):
"""Validate all parameters for the transformer.

Returns
-------
None.

"""
if self.parallel is not None and self.parallel != 'imap_unordered':
raise ValueError('Parallel mode is not set correctly.')

if self.parallel == 'imap_unordered' and self.n_jobs is None:
self.n_jobs = multiprocessing.cpu_count()


def validate_input(self, X):
"""Validate the given input and raise errors if it is invalid.

Parameters
----------
X : list
The input to check. Should be a list of graph.

Raises
------
ValueError
Raise if the input is not correct.

Returns
-------
X : list
The input. A list of graph.

"""
if X is None:
raise ValueError('Please add graphs before computing.')
elif not isinstance(X, list):
raise ValueError('Cannot detect graphs.')
elif len(X) == 0:
raise ValueError('The graph list given is empty. No computation will be performed.')

return X


def compute_kernel_matrix(self, Y=None):
"""Compute the kernel matrix between a given target graphs (Y) and
the fitted graphs (X / self._graphs) or the Gram matrix for the fitted
graphs (X / self._graphs).

Parameters
----------
Y : list of graphs, optional
The target graphs. The default is None. If None kernel is computed
between X and itself.

Returns
-------
kernel_matrix : numpy array, shape = [n_targets, n_inputs]
The computed kernel matrix.

"""
if Y is None:
# Compute Gram matrix for self._graphs (X).
kernel_matrix = self._compute_gram_matrix()
# self._gram_matrix_unnorm = np.copy(self._gram_matrix)

else:
# Compute kernel matrix between Y and self._graphs (X).
start_time = time.time()

if self.parallel == 'imap_unordered':
kernel_matrix = self._compute_kernel_matrix_imap_unordered(Y)

elif self.parallel is None:
kernel_matrix = self._compute_kernel_matrix_series(Y)

self._run_time = time.time() - start_time
if self.verbose:
print('Kernel matrix of size (%d, %d) built in %s seconds.'
% (len(Y), len(self._graphs), self._run_time))

return kernel_matrix


def _compute_kernel_matrix_series(self, Y):
"""Compute the kernel matrix between a given target graphs (Y) and
the fitted graphs (X / self._graphs) without parallelization.

Parameters
----------
Y : list of graphs, optional
The target graphs.

Returns
-------
kernel_matrix : numpy array, shape = [n_targets, n_inputs]
The computed kernel matrix.

"""
kernel_matrix = np.zeros((len(Y), len(self._graphs)))

for i_y, g_y in enumerate(Y):
for i_x, g_x in enumerate(self._graphs):
kernel_matrix[i_y, i_x] = self.pairwise_kernel(g_y, g_x)

return kernel_matrix


def _compute_kernel_matrix_imap_unordered(self, Y):
"""Compute the kernel matrix between a given target graphs (Y) and
the fitted graphs (X / self._graphs) using imap unordered parallelization.

Parameters
----------
Y : list of graphs, optional
The target graphs.

Returns
-------
kernel_matrix : numpy array, shape = [n_targets, n_inputs]
The computed kernel matrix.

"""
raise Exception('Parallelization for kernel matrix is not implemented.')


def diagonals(self):
"""Compute the kernel matrix diagonals of the fit/transformed data.

Returns
-------
X_diag : numpy array
The diagonal of the kernel matrix between the fitted data.
This consists of each element calculated with itself.

Y_diag : numpy array
The diagonal of the kernel matrix, of the transform.
This consists of each element calculated with itself.

"""
# Check if method "fit" had been called.
check_is_fitted(self, ['_graphs'])

# Check if the diagonals of X exist.
try:
check_is_fitted(self, ['_X_diag'])
except NotFittedError:
# Compute diagonals of X.
self._X_diag = np.empty(shape=(len(self._graphs),))
for i, x in enumerate(self._graphs):
self._X_diag[i] = self.pairwise_kernel(x, x) # @todo: parallel?

try:
# If transform has happened, return both diagonals.
check_is_fitted(self, ['_Y'])
self._Y_diag = np.empty(shape=(len(self._Y),))
for (i, y) in enumerate(self._Y):
self._Y_diag[i] = self.pairwise_kernel(y, y) # @todo: parallel?

return self._X_diag, self._Y_diag
except NotFittedError:
# Else just return both X_diag
return self._X_diag


# @abstractmethod
def pairwise_kernel(self, x, y):
"""Compute pairwise kernel between two graphs.

Parameters
----------
x, y : NetworkX Graph.
Graphs bewteen which the kernel is computed.

Returns
-------
kernel: float
The computed kernel.

# Notes
# -----
# This method is abstract and must be implemented by a subclass.

"""
raise NotImplementedError('Pairwise kernel computation is not implemented!')


##########################################################################
# The following is the 2nd paradigm to compute kernel matrix. It is
# simplified and not compatible with `scikit-learn`.
##########################################################################


def compute(self, *graphs, **kwargs):
self._parallel = kwargs.get('parallel', 'imap_unordered')
self._n_jobs = kwargs.get('n_jobs', multiprocessing.cpu_count())
self._normalize = kwargs.get('normalize', True)
self._verbose = kwargs.get('verbose', 2)
self.parallel = kwargs.get('parallel', 'imap_unordered')
self.n_jobs = kwargs.get('n_jobs', multiprocessing.cpu_count())
self.normalize = kwargs.get('normalize', True)
self.verbose = kwargs.get('verbose', 2)
self.validate_parameters()

if len(graphs) == 1:
if not isinstance(graphs[0], list):
@@ -40,7 +385,7 @@ class GraphKernel(object):
self._graphs = [g.copy() for g in graphs[0]] # @todo: might be very slow.
self._gram_matrix = self._compute_gram_matrix()
self._gram_matrix_unnorm = np.copy(self._gram_matrix)
if self._normalize:
if self.normalize:
self._gram_matrix = normalize_gram_matrix(self._gram_matrix)
return self._gram_matrix, self._run_time

@@ -103,15 +448,15 @@ class GraphKernel(object):
def _compute_gram_matrix(self):
start_time = time.time()

if self._parallel == 'imap_unordered':
if self.parallel == 'imap_unordered':
gram_matrix = self._compute_gm_imap_unordered()
elif self._parallel is None:
elif self.parallel is None:
gram_matrix = self._compute_gm_series()
else:
raise Exception('Parallel mode is not set correctly.')

self._run_time = time.time() - start_time
if self._verbose:
if self.verbose:
print('Gram matrix of size %d built in %s seconds.'
% (len(self._graphs), self._run_time))

@@ -129,15 +474,15 @@ class GraphKernel(object):
def _compute_kernel_list(self, g1, g_list):
start_time = time.time()

if self._parallel == 'imap_unordered':
if self.parallel == 'imap_unordered':
kernel_list = self._compute_kernel_list_imap_unordered(g1, g_list)
elif self._parallel is None:
elif self.parallel is None:
kernel_list = self._compute_kernel_list_series(g1, g_list)
else:
raise Exception('Parallel mode is not set correctly.')

self._run_time = time.time() - start_time
if self._verbose:
if self.verbose:
print('Graph kernel bewteen a graph and a list of %d graphs built in %s seconds.'
% (len(g_list), self._run_time))

@@ -158,7 +503,7 @@ class GraphKernel(object):
kernel = self._compute_single_kernel_series(g1, g2)

self._run_time = time.time() - start_time
if self._verbose:
if self.verbose:
print('Graph kernel bewteen two graphs built in %s seconds.' % (self._run_time))

return kernel
@@ -185,24 +530,24 @@ class GraphKernel(object):
return self._graphs


@property
def parallel(self):
return self._parallel
# @property
# def parallel(self):
# return self.parallel


@property
def n_jobs(self):
return self._n_jobs
# @property
# def n_jobs(self):
# return self.n_jobs


@property
def verbose(self):
return self._verbose
# @property
# def verbose(self):
# return self.verbose


@property
def normalize(self):
return self._normalize
# @property
# def normalize(self):
# return self.normalize


@property


+ 14
- 14
gklearn/kernels/marginalized.py View File

@@ -46,7 +46,7 @@ class Marginalized(GraphKernel):
self._add_dummy_labels(self._graphs)

if self._remove_totters:
iterator = get_iters(self._graphs, desc='removing tottering', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='removing tottering', file=sys.stdout, verbose=(self.verbose >= 2))
# @todo: this may not work.
self._graphs = [untotterTransformation(G, self._node_labels, self._edge_labels) for G in iterator]

@@ -57,7 +57,7 @@ class Marginalized(GraphKernel):
itr = combinations_with_replacement(range(0, len(self._graphs)), 2)
len_itr = int(len(self._graphs) * (len(self._graphs) + 1) / 2)
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout,
length=len_itr, verbose=(self._verbose >= 2))
length=len_itr, verbose=(self.verbose >= 2))
for i, j in iterator:
kernel = self._kernel_do(self._graphs[i], self._graphs[j])
gram_matrix[i][j] = kernel
@@ -70,16 +70,16 @@ class Marginalized(GraphKernel):
self._add_dummy_labels(self._graphs)

if self._remove_totters:
pool = Pool(self._n_jobs)
pool = Pool(self.n_jobs)
itr = range(0, len(self._graphs))
if len(self._graphs) < 100 * self._n_jobs:
chunksize = int(len(self._graphs) / self._n_jobs) + 1
if len(self._graphs) < 100 * self.n_jobs:
chunksize = int(len(self._graphs) / self.n_jobs) + 1
else:
chunksize = 100
remove_fun = self._wrapper_untotter
iterator = get_iters(pool.imap_unordered(remove_fun, itr, chunksize),
desc='removing tottering', file=sys.stdout,
length=len(self._graphs), verbose=(self._verbose >= 2))
length=len(self._graphs), verbose=(self.verbose >= 2))
for i, g in iterator:
self._graphs[i] = g
pool.close()
@@ -93,7 +93,7 @@ class Marginalized(GraphKernel):
G_gn = gn_toshare
do_fun = self._wrapper_kernel_do
parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=init_worker,
glbv=(self._graphs,), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(self._graphs,), n_jobs=self.n_jobs, verbose=self.verbose)

return gram_matrix

@@ -103,13 +103,13 @@ class Marginalized(GraphKernel):

if self._remove_totters:
g1 = untotterTransformation(g1, self._node_labels, self._edge_labels) # @todo: this may not work.
iterator = get_iters(g_list, desc='removing tottering', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(g_list, desc='removing tottering', file=sys.stdout, verbose=(self.verbose >= 2))
# @todo: this may not work.
g_list = [untotterTransformation(G, self._node_labels, self._edge_labels) for G in iterator]

# compute kernel list.
kernel_list = [None] * len(g_list)
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self._verbose >= 2))
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self.verbose >= 2))
for i in iterator:
kernel = self._kernel_do(g1, g_list[i])
kernel_list[i] = kernel
@@ -122,16 +122,16 @@ class Marginalized(GraphKernel):

if self._remove_totters:
g1 = untotterTransformation(g1, self._node_labels, self._edge_labels) # @todo: this may not work.
pool = Pool(self._n_jobs)
pool = Pool(self.n_jobs)
itr = range(0, len(g_list))
if len(g_list) < 100 * self._n_jobs:
chunksize = int(len(g_list) / self._n_jobs) + 1
if len(g_list) < 100 * self.n_jobs:
chunksize = int(len(g_list) / self.n_jobs) + 1
else:
chunksize = 100
remove_fun = self._wrapper_untotter
iterator = get_iters(pool.imap_unordered(remove_fun, itr, chunksize),
desc='removing tottering', file=sys.stdout,
length=len(g_list), verbose=(self._verbose >= 2))
length=len(g_list), verbose=(self.verbose >= 2))
for i, g in iterator:
g_list[i] = g
pool.close()
@@ -151,7 +151,7 @@ class Marginalized(GraphKernel):
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=init_worker, glbv=(g1, g_list), method='imap_unordered',
n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)

return kernel_list



+ 14
- 14
gklearn/kernels/path_up_to_h.py View File

@@ -41,10 +41,10 @@ class PathUpToH(GraphKernel): # @todo: add function for k_func is None

from itertools import combinations_with_replacement
itr_kernel = combinations_with_replacement(range(0, len(self._graphs)), 2)
iterator_ps = get_iters(range(0, len(self._graphs)), desc='getting paths', file=sys.stdout, length=len(self._graphs), verbose=(self._verbose >= 2))
iterator_ps = get_iters(range(0, len(self._graphs)), desc='getting paths', file=sys.stdout, length=len(self._graphs), verbose=(self.verbose >= 2))
len_itr = int(len(self._graphs) * (len(self._graphs) + 1) / 2)
iterator_kernel = get_iters(itr_kernel, desc='Computing kernels',
file=sys.stdout, length=len_itr, verbose=(self._verbose >= 2))
file=sys.stdout, length=len_itr, verbose=(self.verbose >= 2))

gram_matrix = np.zeros((len(self._graphs), len(self._graphs)))

@@ -69,10 +69,10 @@ class PathUpToH(GraphKernel): # @todo: add function for k_func is None

# get all paths of all graphs before computing kernels to save time,
# but this may cost a lot of memory for large datasets.
pool = Pool(self._n_jobs)
pool = Pool(self.n_jobs)
itr = zip(self._graphs, range(0, len(self._graphs)))
if len(self._graphs) < 100 * self._n_jobs:
chunksize = int(len(self._graphs) / self._n_jobs) + 1
if len(self._graphs) < 100 * self.n_jobs:
chunksize = int(len(self._graphs) / self.n_jobs) + 1
else:
chunksize = 100
all_paths = [[] for _ in range(len(self._graphs))]
@@ -84,7 +84,7 @@ class PathUpToH(GraphKernel): # @todo: add function for k_func is None
get_ps_fun = partial(self._wrapper_find_all_paths_until_length, False)
iterator = get_iters(pool.imap_unordered(get_ps_fun, itr, chunksize),
desc='getting paths', file=sys.stdout,
length=len(self._graphs), verbose=(self._verbose >= 2))
length=len(self._graphs), verbose=(self.verbose >= 2))
for i, ps in iterator:
all_paths[i] = ps
pool.close()
@@ -109,7 +109,7 @@ class PathUpToH(GraphKernel): # @todo: add function for k_func is None
G_plist = plist_toshare
do_fun = self._wrapper_kernel_do_kernelless # @todo: what is this?
parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=init_worker,
glbv=(all_paths,), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(all_paths,), n_jobs=self.n_jobs, verbose=self.verbose)

return gram_matrix

@@ -117,8 +117,8 @@ class PathUpToH(GraphKernel): # @todo: add function for k_func is None
def _compute_kernel_list_series(self, g1, g_list):
self._add_dummy_labels(g_list + [g1])

iterator_ps = get_iters(g_list, desc='getting paths', file=sys.stdout, verbose=(self._verbose >= 2))
iterator_kernel = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self._verbose >= 2))
iterator_ps = get_iters(g_list, desc='getting paths', file=sys.stdout, verbose=(self.verbose >= 2))
iterator_kernel = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self.verbose >= 2))

kernel_list = [None] * len(g_list)

@@ -143,10 +143,10 @@ class PathUpToH(GraphKernel): # @todo: add function for k_func is None

# get all paths of all graphs before computing kernels to save time,
# but this may cost a lot of memory for large datasets.
pool = Pool(self._n_jobs)
pool = Pool(self.n_jobs)
itr = zip(g_list, range(0, len(g_list)))
if len(g_list) < 100 * self._n_jobs:
chunksize = int(len(g_list) / self._n_jobs) + 1
if len(g_list) < 100 * self.n_jobs:
chunksize = int(len(g_list) / self.n_jobs) + 1
else:
chunksize = 100
paths_g_list = [[] for _ in range(len(g_list))]
@@ -161,7 +161,7 @@ class PathUpToH(GraphKernel): # @todo: add function for k_func is None
get_ps_fun = partial(self._wrapper_find_all_paths_until_length, False)
iterator = get_iters(pool.imap_unordered(get_ps_fun, itr, chunksize),
desc='getting paths', file=sys.stdout,
length=len(g_list), verbose=(self._verbose >= 2))
length=len(g_list), verbose=(self.verbose >= 2))
for i, ps in iterator:
paths_g_list[i] = ps
pool.close()
@@ -180,7 +180,7 @@ class PathUpToH(GraphKernel): # @todo: add function for k_func is None
itr = range(len(g_list))
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=init_worker, glbv=(paths_g1, paths_g_list), method='imap_unordered', n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
init_worker=init_worker, glbv=(paths_g1, paths_g_list), method='imap_unordered', n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)

return kernel_list



+ 14
- 14
gklearn/kernels/shortest_path.py View File

@@ -38,7 +38,7 @@ class ShortestPath(GraphKernel):
def _compute_gm_series(self):
self._all_graphs_have_edges(self._graphs)
# get shortest path graph of each graph.
iterator = get_iters(self._graphs, desc='getting sp graphs', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='getting sp graphs', file=sys.stdout, verbose=(self.verbose >= 2))
self._graphs = [getSPGraph(g, edge_weight=self._edge_weight) for g in iterator]

# compute Gram matrix.
@@ -48,7 +48,7 @@ class ShortestPath(GraphKernel):
itr = combinations_with_replacement(range(0, len(self._graphs)), 2)
len_itr = int(len(self._graphs) * (len(self._graphs) + 1) / 2)
iterator = get_iters(itr, desc='Computing kernels',
length=len_itr, file=sys.stdout,verbose=(self._verbose >= 2))
length=len_itr, file=sys.stdout,verbose=(self.verbose >= 2))
for i, j in iterator:
kernel = self._sp_do(self._graphs[i], self._graphs[j])
gram_matrix[i][j] = kernel
@@ -60,16 +60,16 @@ class ShortestPath(GraphKernel):
def _compute_gm_imap_unordered(self):
self._all_graphs_have_edges(self._graphs)
# get shortest path graph of each graph.
pool = Pool(self._n_jobs)
pool = Pool(self.n_jobs)
get_sp_graphs_fun = self._wrapper_get_sp_graphs
itr = zip(self._graphs, range(0, len(self._graphs)))
if len(self._graphs) < 100 * self._n_jobs:
chunksize = int(len(self._graphs) / self._n_jobs) + 1
if len(self._graphs) < 100 * self.n_jobs:
chunksize = int(len(self._graphs) / self.n_jobs) + 1
else:
chunksize = 100
iterator = get_iters(pool.imap_unordered(get_sp_graphs_fun, itr, chunksize),
desc='getting sp graphs', file=sys.stdout,
length=len(self._graphs), verbose=(self._verbose >= 2))
length=len(self._graphs), verbose=(self.verbose >= 2))
for i, g in iterator:
self._graphs[i] = g
pool.close()
@@ -83,7 +83,7 @@ class ShortestPath(GraphKernel):
G_gs = gs_toshare
do_fun = self._wrapper_sp_do
parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=init_worker,
glbv=(self._graphs,), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(self._graphs,), n_jobs=self.n_jobs, verbose=self.verbose)

return gram_matrix

@@ -92,12 +92,12 @@ class ShortestPath(GraphKernel):
self._all_graphs_have_edges([g1] + g_list)
# get shortest path graphs of g1 and each graph in g_list.
g1 = getSPGraph(g1, edge_weight=self._edge_weight)
iterator = get_iters(g_list, desc='getting sp graphs', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(g_list, desc='getting sp graphs', file=sys.stdout, verbose=(self.verbose >= 2))
g_list = [getSPGraph(g, edge_weight=self._edge_weight) for g in iterator]

# compute kernel list.
kernel_list = [None] * len(g_list)
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self._verbose >= 2))
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self.verbose >= 2))
for i in iterator:
kernel = self._sp_do(g1, g_list[i])
kernel_list[i] = kernel
@@ -109,16 +109,16 @@ class ShortestPath(GraphKernel):
self._all_graphs_have_edges([g1] + g_list)
# get shortest path graphs of g1 and each graph in g_list.
g1 = getSPGraph(g1, edge_weight=self._edge_weight)
pool = Pool(self._n_jobs)
pool = Pool(self.n_jobs)
get_sp_graphs_fun = self._wrapper_get_sp_graphs
itr = zip(g_list, range(0, len(g_list)))
if len(g_list) < 100 * self._n_jobs:
chunksize = int(len(g_list) / self._n_jobs) + 1
if len(g_list) < 100 * self.n_jobs:
chunksize = int(len(g_list) / self.n_jobs) + 1
else:
chunksize = 100
iterator = get_iters(pool.imap_unordered(get_sp_graphs_fun, itr, chunksize),
desc='getting sp graphs', file=sys.stdout,
length=len(g_list), verbose=(self._verbose >= 2))
length=len(g_list), verbose=(self.verbose >= 2))
for i, g in iterator:
g_list[i] = g
pool.close()
@@ -137,7 +137,7 @@ class ShortestPath(GraphKernel):
itr = range(len(g_list))
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=init_worker, glbv=(g1, g_list), method='imap_unordered', n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
init_worker=init_worker, glbv=(g1, g_list), method='imap_unordered', n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)

return kernel_list



+ 19
- 19
gklearn/kernels/spectral_decomposition.py View File

@@ -28,9 +28,9 @@ class SpectralDecomposition(RandomWalkMeta):


def _compute_gm_series(self):
self._check_edge_weight(self._graphs, self._verbose)
self._check_edge_weight(self._graphs, self.verbose)
self._check_graphs(self._graphs)
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('All labels are ignored. Only works for undirected graphs.')

@@ -41,7 +41,7 @@ class SpectralDecomposition(RandomWalkMeta):
# precompute the spectral decomposition of each graph.
P_list = []
D_list = []
iterator = get_iters(self._graphs, desc='spectral decompose', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='spectral decompose', file=sys.stdout, verbose=(self.verbose >= 2))
for G in iterator:
# don't normalize adjacency matrices if q is a uniform vector. Note
# A actually is the transpose of the adjacency matrix.
@@ -58,7 +58,7 @@ class SpectralDecomposition(RandomWalkMeta):
from itertools import combinations_with_replacement
itr = combinations_with_replacement(range(0, len(self._graphs)), 2)
len_itr = int(len(self._graphs) * (len(self._graphs) + 1) / 2)
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout, length=len_itr, verbose=(self._verbose >= 2))
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout, length=len_itr, verbose=(self.verbose >= 2))

for i, j in iterator:
kernel = self._kernel_do(q_T_list[i], q_T_list[j], P_list[i], P_list[j], D_list[i], D_list[j], self._weight, self._sub_kernel)
@@ -74,9 +74,9 @@ class SpectralDecomposition(RandomWalkMeta):


def _compute_gm_imap_unordered(self):
self._check_edge_weight(self._graphs, self._verbose)
self._check_edge_weight(self._graphs, self.verbose)
self._check_graphs(self._graphs)
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('All labels are ignored. Only works for undirected graphs.')

@@ -87,7 +87,7 @@ class SpectralDecomposition(RandomWalkMeta):
# precompute the spectral decomposition of each graph.
P_list = []
D_list = []
iterator = get_iters(self._graphs, desc='spectral decompose', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='spectral decompose', file=sys.stdout, verbose=(self.verbose >= 2))
for G in iterator:
# don't normalize adjacency matrices if q is a uniform vector. Note
# A actually is the transpose of the adjacency matrix.
@@ -107,7 +107,7 @@ class SpectralDecomposition(RandomWalkMeta):

do_fun = self._wrapper_kernel_do
parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=init_worker,
glbv=(q_T_list, P_list, D_list), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(q_T_list, P_list, D_list), n_jobs=self.n_jobs, verbose=self.verbose)

else: # @todo
pass
@@ -118,9 +118,9 @@ class SpectralDecomposition(RandomWalkMeta):


def _compute_kernel_list_series(self, g1, g_list):
self._check_edge_weight(g_list + [g1], self._verbose)
self._check_edge_weight(g_list + [g1], self.verbose)
self._check_graphs(g_list + [g1])
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('All labels are ignored. Only works for undirected graphs.')

@@ -133,7 +133,7 @@ class SpectralDecomposition(RandomWalkMeta):
D1, P1 = np.linalg.eig(A1)
P_list = []
D_list = []
iterator = get_iters(g_list, desc='spectral decompose', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(g_list, desc='spectral decompose', file=sys.stdout, verbose=(self.verbose >= 2))
for G in iterator:
# don't normalize adjacency matrices if q is a uniform vector. Note
# A actually is the transpose of the adjacency matrix.
@@ -145,7 +145,7 @@ class SpectralDecomposition(RandomWalkMeta):
if self._p is None: # p is uniform distribution as default.
q_T1 = 1 / nx.number_of_nodes(g1)
q_T_list = [np.full((1, nx.number_of_nodes(G)), 1 / nx.number_of_nodes(G)) for G in g_list]
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self._verbose >= 2))
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self.verbose >= 2))

for i in iterator:
kernel = self._kernel_do(q_T1, q_T_list[i], P1, P_list[i], D1, D_list[i], self._weight, self._sub_kernel)
@@ -160,9 +160,9 @@ class SpectralDecomposition(RandomWalkMeta):


def _compute_kernel_list_imap_unordered(self, g1, g_list):
self._check_edge_weight(g_list + [g1], self._verbose)
self._check_edge_weight(g_list + [g1], self.verbose)
self._check_graphs(g_list + [g1])
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('All labels are ignored. Only works for undirected graphs.')

@@ -175,8 +175,8 @@ class SpectralDecomposition(RandomWalkMeta):
D1, P1 = np.linalg.eig(A1)
P_list = []
D_list = []
if self._verbose >= 2:
iterator = tqdm(g_list, desc='spectral decompose', file=sys.stdout)
if self.verbose >= 2:
iterator = get_iters(g_list, desc='spectral decompose', file=sys.stdout)
else:
iterator = g_list
for G in iterator:
@@ -207,7 +207,7 @@ class SpectralDecomposition(RandomWalkMeta):
itr = range(len(g_list))
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=init_worker, glbv=(q_T1, P1, D1, q_T_list, P_list, D_list), method='imap_unordered', n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
init_worker=init_worker, glbv=(q_T1, P1, D1, q_T_list, P_list, D_list), method='imap_unordered', n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)

else: # @todo
pass
@@ -222,9 +222,9 @@ class SpectralDecomposition(RandomWalkMeta):


def _compute_single_kernel_series(self, g1, g2):
self._check_edge_weight([g1] + [g2], self._verbose)
self._check_edge_weight([g1] + [g2], self.verbose)
self._check_graphs([g1] + [g2])
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('All labels are ignored. Only works for undirected graphs.')



+ 14
- 14
gklearn/kernels/structural_sp.py View File

@@ -41,7 +41,7 @@ class StructuralSP(GraphKernel):
def _compute_gm_series(self):
# get shortest paths of each graph in the graphs.
splist = []
iterator = get_iters(self._graphs, desc='getting sp graphs', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='getting sp graphs', file=sys.stdout, verbose=(self.verbose >= 2))
if self._compute_method == 'trie':
for g in iterator:
splist.append(self._get_sps_as_trie(g))
@@ -56,7 +56,7 @@ class StructuralSP(GraphKernel):
itr = combinations_with_replacement(range(0, len(self._graphs)), 2)
len_itr = int(len(self._graphs) * (len(self._graphs) + 1) / 2)
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout,
length=len_itr, verbose=(self._verbose >= 2))
length=len_itr, verbose=(self.verbose >= 2))
if self._compute_method == 'trie':
for i, j in iterator:
kernel = self._ssp_do_trie(self._graphs[i], self._graphs[j], splist[i], splist[j])
@@ -76,10 +76,10 @@ class StructuralSP(GraphKernel):
def _compute_gm_imap_unordered(self):
# get shortest paths of each graph in the graphs.
splist = [None] * len(self._graphs)
pool = Pool(self._n_jobs)
pool = Pool(self.n_jobs)
itr = zip(self._graphs, range(0, len(self._graphs)))
if len(self._graphs) < 100 * self._n_jobs:
chunksize = int(len(self._graphs) / self._n_jobs) + 1
if len(self._graphs) < 100 * self.n_jobs:
chunksize = int(len(self._graphs) / self.n_jobs) + 1
else:
chunksize = 100
# get shortest path graphs of self._graphs
@@ -89,7 +89,7 @@ class StructuralSP(GraphKernel):
get_sps_fun = self._wrapper_get_sps_naive
iterator = get_iters(pool.imap_unordered(get_sps_fun, itr, chunksize),
desc='getting shortest paths', file=sys.stdout,
length=len(self._graphs), verbose=(self._verbose >= 2))
length=len(self._graphs), verbose=(self.verbose >= 2))
for i, sp in iterator:
splist[i] = sp
pool.close()
@@ -107,7 +107,7 @@ class StructuralSP(GraphKernel):
else:
do_fun = self._wrapper_ssp_do_naive
parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=init_worker,
glbv=(splist, self._graphs), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(splist, self._graphs), n_jobs=self.n_jobs, verbose=self.verbose)

return gram_matrix

@@ -117,7 +117,7 @@ class StructuralSP(GraphKernel):
sp1 = get_shortest_paths(g1, self._edge_weight, self._ds_infos['directed'])
splist = []
iterator = get_iters(g_list, desc='getting sp graphs', file=sys.stdout,
verbose=(self._verbose >= 2))
verbose=(self.verbose >= 2))
if self._compute_method == 'trie':
for g in iterator:
splist.append(self._get_sps_as_trie(g))
@@ -128,7 +128,7 @@ class StructuralSP(GraphKernel):
# compute kernel list.
kernel_list = [None] * len(g_list)
iterator = get_iters(range(len(g_list)), desc='Computing kernels',
file=sys.stdout, length=len(g_list), verbose=(self._verbose >= 2))
file=sys.stdout, length=len(g_list), verbose=(self.verbose >= 2))
if self._compute_method == 'trie':
for i in iterator:
kernel = self._ssp_do_trie(g1, g_list[i], sp1, splist[i])
@@ -145,10 +145,10 @@ class StructuralSP(GraphKernel):
# get shortest paths of g1 and each graph in g_list.
sp1 = get_shortest_paths(g1, self._edge_weight, self._ds_infos['directed'])
splist = [None] * len(g_list)
pool = Pool(self._n_jobs)
pool = Pool(self.n_jobs)
itr = zip(g_list, range(0, len(g_list)))
if len(g_list) < 100 * self._n_jobs:
chunksize = int(len(g_list) / self._n_jobs) + 1
if len(g_list) < 100 * self.n_jobs:
chunksize = int(len(g_list) / self.n_jobs) + 1
else:
chunksize = 100
# get shortest path graphs of g_list
@@ -158,7 +158,7 @@ class StructuralSP(GraphKernel):
get_sps_fun = self._wrapper_get_sps_naive
iterator = get_iters(pool.imap_unordered(get_sps_fun, itr, chunksize),
desc='getting shortest paths', file=sys.stdout,
length=len(g_list), verbose=(self._verbose >= 2))
length=len(g_list), verbose=(self.verbose >= 2))
for i, sp in iterator:
splist[i] = sp
pool.close()
@@ -182,7 +182,7 @@ class StructuralSP(GraphKernel):
itr = range(len(g_list))
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=init_worker, glbv=(sp1, splist, g1, g_list), method='imap_unordered', n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
init_worker=init_worker, glbv=(sp1, splist, g1, g_list), method='imap_unordered', n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)

return kernel_list



+ 18
- 18
gklearn/kernels/sylvester_equation.py View File

@@ -27,9 +27,9 @@ class SylvesterEquation(RandomWalkMeta):


def _compute_gm_series(self):
self._check_edge_weight(self._graphs, self._verbose)
self._check_edge_weight(self._graphs, self.verbose)
self._check_graphs(self._graphs)
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('All labels are ignored.')

@@ -41,7 +41,7 @@ class SylvesterEquation(RandomWalkMeta):
if self._q is None:
# don't normalize adjacency matrices if q is a uniform vector. Note
# A_wave_list actually contains the transposes of the adjacency matrices.
iterator = get_iters(self._graphs, desc='compute adjacency matrices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='compute adjacency matrices', file=sys.stdout, verbose=(self.verbose >= 2))
A_wave_list = [nx.adjacency_matrix(G, self._edge_weight).todense().transpose() for G in iterator]
# # normalized adjacency matrices
# A_wave_list = []
@@ -55,7 +55,7 @@ class SylvesterEquation(RandomWalkMeta):
from itertools import combinations_with_replacement
itr = combinations_with_replacement(range(0, len(self._graphs)), 2)
len_itr = int(len(self._graphs) * (len(self._graphs) + 1) / 2)
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout, length=len_itr, verbose=(self._verbose >= 2))
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout, length=len_itr, verbose=(self.verbose >= 2))

for i, j in iterator:
kernel = self._kernel_do(A_wave_list[i], A_wave_list[j], lmda)
@@ -71,9 +71,9 @@ class SylvesterEquation(RandomWalkMeta):


def _compute_gm_imap_unordered(self):
self._check_edge_weight(self._graphs, self._verbose)
self._check_edge_weight(self._graphs, self.verbose)
self._check_graphs(self._graphs)
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('All labels are ignored.')

@@ -83,7 +83,7 @@ class SylvesterEquation(RandomWalkMeta):
if self._q is None:
# don't normalize adjacency matrices if q is a uniform vector. Note
# A_wave_list actually contains the transposes of the adjacency matrices.
iterator = get_iters(self._graphs, desc='compute adjacency matrices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(self._graphs, desc='compute adjacency matrices', file=sys.stdout, verbose=(self.verbose >= 2))
A_wave_list = [nx.adjacency_matrix(G, self._edge_weight).todense().transpose() for G in iterator] # @todo: parallel?

if self._p is None: # p is uniform distribution as default.
@@ -94,7 +94,7 @@ class SylvesterEquation(RandomWalkMeta):
do_fun = self._wrapper_kernel_do

parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=init_worker,
glbv=(A_wave_list,), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(A_wave_list,), n_jobs=self.n_jobs, verbose=self.verbose)

else: # @todo
pass
@@ -105,9 +105,9 @@ class SylvesterEquation(RandomWalkMeta):


def _compute_kernel_list_series(self, g1, g_list):
self._check_edge_weight(g_list + [g1], self._verbose)
self._check_edge_weight(g_list + [g1], self.verbose)
self._check_graphs(g_list + [g1])
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('All labels are ignored.')

@@ -120,11 +120,11 @@ class SylvesterEquation(RandomWalkMeta):
# don't normalize adjacency matrices if q is a uniform vector. Note
# A_wave_list actually contains the transposes of the adjacency matrices.
A_wave_1 = nx.adjacency_matrix(g1, self._edge_weight).todense().transpose()
iterator = get_iters(g_list, desc='compute adjacency matrices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(g_list, desc='compute adjacency matrices', file=sys.stdout, verbose=(self.verbose >= 2))
A_wave_list = [nx.adjacency_matrix(G, self._edge_weight).todense().transpose() for G in iterator]

if self._p is None: # p is uniform distribution as default.
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self._verbose >= 2))
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self.verbose >= 2))

for i in iterator:
kernel = self._kernel_do(A_wave_1, A_wave_list[i], lmda)
@@ -139,9 +139,9 @@ class SylvesterEquation(RandomWalkMeta):


def _compute_kernel_list_imap_unordered(self, g1, g_list):
self._check_edge_weight(g_list + [g1], self._verbose)
self._check_edge_weight(g_list + [g1], self.verbose)
self._check_graphs(g_list + [g1])
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('All labels are ignored.')

@@ -152,7 +152,7 @@ class SylvesterEquation(RandomWalkMeta):
# don't normalize adjacency matrices if q is a uniform vector. Note
# A_wave_list actually contains the transposes of the adjacency matrices.
A_wave_1 = nx.adjacency_matrix(g1, self._edge_weight).todense().transpose()
iterator = get_iters(g_list, desc='compute adjacency matrices', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(g_list, desc='compute adjacency matrices', file=sys.stdout, verbose=(self.verbose >= 2))
A_wave_list = [nx.adjacency_matrix(G, self._edge_weight).todense().transpose() for G in iterator] # @todo: parallel?

if self._p is None: # p is uniform distribution as default.
@@ -169,7 +169,7 @@ class SylvesterEquation(RandomWalkMeta):
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=init_worker, glbv=(A_wave_1, A_wave_list), method='imap_unordered',
n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)

else: # @todo
pass
@@ -184,9 +184,9 @@ class SylvesterEquation(RandomWalkMeta):


def _compute_single_kernel_series(self, g1, g2):
self._check_edge_weight([g1] + [g2], self._verbose)
self._check_edge_weight([g1] + [g2], self.verbose)
self._check_graphs([g1] + [g2])
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('All labels are ignored.')



+ 313
- 87
gklearn/kernels/treelet.py View File

@@ -18,6 +18,8 @@ import numpy as np
import networkx as nx
from collections import Counter
from itertools import chain
from sklearn.utils.validation import check_is_fitted
from sklearn.exceptions import NotFittedError
from gklearn.utils import SpecialLabel
from gklearn.utils.parallel import parallel_gm, parallel_me
from gklearn.utils.utils import find_all_paths, get_mlti_dim_node_attrs
@@ -26,14 +28,211 @@ from gklearn.kernels import GraphKernel

class Treelet(GraphKernel):

def __init__(self, **kwargs):
GraphKernel.__init__(self)
self._node_labels = kwargs.get('node_labels', [])
self._edge_labels = kwargs.get('edge_labels', [])
self._sub_kernel = kwargs.get('sub_kernel', None)
self._ds_infos = kwargs.get('ds_infos', {})
if self._sub_kernel is None:
raise Exception('Sub kernel not set.')
def __init__(self, parallel=None, n_jobs=None, chunksize=None, normalize=True, verbose=2, precompute_canonkeys=True, save_canonkeys=False, **kwargs):
"""Initialise a treelet kernel.
"""
super().__init__(parallel=parallel, n_jobs=n_jobs, chunksize=chunksize, normalize=normalize, verbose=verbose)
self.node_labels = kwargs.get('node_labels', [])
self.edge_labels = kwargs.get('edge_labels', [])
self.sub_kernel = kwargs.get('sub_kernel', None)
self.ds_infos = kwargs.get('ds_infos', {})
self.precompute_canonkeys = precompute_canonkeys
self.save_canonkeys = save_canonkeys


##########################################################################
# The following is the 1st paradigm to compute kernel matrix, which is
# compatible with `scikit-learn`.
# -------------------------------------------------------------------
# Special thanks to the "GraKeL" library for providing an excellent template!
##########################################################################


def clear_attributes(self):
super().clear_attributes()
if hasattr(self, '_canonkeys'):
delattr(self, '_canonkeys')
if hasattr(self, '_Y_canonkeys'):
delattr(self, '_Y_canonkeys')
if hasattr(self, '_dummy_labels_considered'):
delattr(self, '_dummy_labels_considered')


def validate_parameters(self):
"""Validate all parameters for the transformer.

Returns
-------
None.

"""
super().validate_parameters()
if self.sub_kernel is None:
raise ValueError('Sub-kernel not set.')


def _compute_kernel_matrix_series(self, Y):
"""Compute the kernel matrix between a given target graphs (Y) and
the fitted graphs (X / self._graphs) without parallelization.

Parameters
----------
Y : list of graphs, optional
The target graphs.

Returns
-------
kernel_matrix : numpy array, shape = [n_targets, n_inputs]
The computed kernel matrix.

"""

# self._add_dummy_labels will modify the input in place.
self._add_dummy_labels() # For self._graphs
# Y = [g.copy() for g in Y] # @todo: ?
self._add_dummy_labels(Y)

# get all canonical keys of all graphs before computing kernels to save
# time, but this may cost a lot of memory for large dataset.

# Canonical keys for self._graphs.
try:
check_is_fitted(self, ['_canonkeys'])
canonkeys_list1 = self._canonkeys
except NotFittedError:
canonkeys_list1 = []
iterator = get_iters(self._graphs, desc='getting canonkeys for X', file=sys.stdout, verbose=(self.verbose >= 2))
for g in iterator:
canonkeys_list1.append(self._get_canonkeys(g))

if self.save_canonkeys:
self._canonkeys = canonkeys_list1

# Canonical keys for Y.
canonkeys_list2 = []
iterator = get_iters(Y, desc='getting canonkeys for Y', file=sys.stdout, verbose=(self.verbose >= 2))
for g in iterator:
canonkeys_list2.append(self._get_canonkeys(g))

if self.save_canonkeys:
self._Y_canonkeys = canonkeys_list2

# compute kernel matrix.
kernel_matrix = np.zeros((len(Y), len(canonkeys_list1)))

from itertools import product
itr = product(range(len(Y)), range(len(canonkeys_list1)))
len_itr = int(len(Y) * len(canonkeys_list1))
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout,
length=len_itr, verbose=(self.verbose >= 2))
for i_y, i_x in iterator:
kernel = self._kernel_do(canonkeys_list2[i_y], canonkeys_list1[i_x])
kernel_matrix[i_y][i_x] = kernel

return kernel_matrix


def _compute_kernel_matrix_imap_unordered(self, Y):
"""Compute the kernel matrix between a given target graphs (Y) and
the fitted graphs (X / self._graphs) using imap unordered parallelization.

Parameters
----------
Y : list of graphs, optional
The target graphs.

Returns
-------
kernel_matrix : numpy array, shape = [n_targets, n_inputs]
The computed kernel matrix.

"""
raise Exception('Parallelization for kernel matrix is not implemented.')


def pairwise_kernel(self, x, y, are_keys=False):
"""Compute pairwise kernel between two graphs.

Parameters
----------
x, y : NetworkX Graph.
Graphs bewteen which the kernel is computed.

are_keys : boolean, optional
If `True`, `x` and `y` are canonical keys, otherwise are graphs.
The default is False.

Returns
-------
kernel: float
The computed kernel.

"""
if are_keys:
# x, y are canonical keys.
kernel = self._kernel_do(x, y)

else:
# x, y are graphs.
kernel = self._compute_single_kernel_series(x, y)

return kernel


def diagonals(self):
"""Compute the kernel matrix diagonals of the fit/transformed data.

Returns
-------
X_diag : numpy array
The diagonal of the kernel matrix between the fitted data.
This consists of each element calculated with itself.

Y_diag : numpy array
The diagonal of the kernel matrix, of the transform.
This consists of each element calculated with itself.

"""
# Check if method "fit" had been called.
check_is_fitted(self, ['_graphs'])

# Check if the diagonals of X exist.
try:
check_is_fitted(self, ['_X_diag'])
except NotFittedError:
# Compute diagonals of X.
self._X_diag = np.empty(shape=(len(self._graphs),))
try:
check_is_fitted(self, ['_canonkeys'])
for i, x in enumerate(self._canonkeys):
self._X_diag[i] = self.pairwise_kernel(x, x, are_keys=True) # @todo: parallel?
except NotFittedError:
for i, x in enumerate(self._graphs):
self._X_diag[i] = self.pairwise_kernel(x, x, are_keys=False) # @todo: parallel?

try:
# If transform has happened, return both diagonals.
check_is_fitted(self, ['_Y'])
self._Y_diag = np.empty(shape=(len(self._Y),))
try:
check_is_fitted(self, ['_Y_canonkeys'])
for (i, y) in enumerate(self._Y_canonkeys):
self._Y_diag[i] = self.pairwise_kernel(y, y, are_keys=True) # @todo: parallel?
except NotFittedError:
for (i, y) in enumerate(self._Y):
self._Y_diag[i] = self.pairwise_kernel(y, y, are_keys=False) # @todo: parallel?

return self._X_diag, self._Y_diag

except NotFittedError:
# Else just return both X_diag
return self._X_diag


##########################################################################
# The following is the 2nd paradigm to compute kernel matrix. It is
# simplified and not compatible with `scikit-learn`.
##########################################################################


def _compute_gm_series(self):
@@ -43,10 +242,13 @@ class Treelet(GraphKernel):
# time, but this may cost a lot of memory for large dataset.
canonkeys = []
iterator = get_iters(self._graphs, desc='getting canonkeys', file=sys.stdout,
verbose=(self._verbose >= 2))
verbose=(self.verbose >= 2))
for g in iterator:
canonkeys.append(self._get_canonkeys(g))

if self.save_canonkeys:
self._canonkeys = canonkeys

# compute Gram matrix.
gram_matrix = np.zeros((len(self._graphs), len(self._graphs)))

@@ -54,7 +256,7 @@ class Treelet(GraphKernel):
itr = combinations_with_replacement(range(0, len(self._graphs)), 2)
len_itr = int(len(self._graphs) * (len(self._graphs) + 1) / 2)
iterator = get_iters(itr, desc='Computing kernels', file=sys.stdout,
length=len_itr, verbose=(self._verbose >= 2))
length=len_itr, verbose=(self.verbose >= 2))
for i, j in iterator:
kernel = self._kernel_do(canonkeys[i], canonkeys[j])
gram_matrix[i][j] = kernel
@@ -68,22 +270,25 @@ class Treelet(GraphKernel):

# get all canonical keys of all graphs before computing kernels to save
# time, but this may cost a lot of memory for large dataset.
pool = Pool(self._n_jobs)
pool = Pool(self.n_jobs)
itr = zip(self._graphs, range(0, len(self._graphs)))
if len(self._graphs) < 100 * self._n_jobs:
chunksize = int(len(self._graphs) / self._n_jobs) + 1
if len(self._graphs) < 100 * self.n_jobs:
chunksize = int(len(self._graphs) / self.n_jobs) + 1
else:
chunksize = 100
canonkeys = [[] for _ in range(len(self._graphs))]
get_fun = self._wrapper_get_canonkeys
iterator = get_iters(pool.imap_unordered(get_fun, itr, chunksize),
desc='getting canonkeys', file=sys.stdout,
length=len(self._graphs), verbose=(self._verbose >= 2))
length=len(self._graphs), verbose=(self.verbose >= 2))
for i, ck in iterator:
canonkeys[i] = ck
pool.close()
pool.join()

if self.save_canonkeys:
self._canonkeys = canonkeys

# compute Gram matrix.
gram_matrix = np.zeros((len(self._graphs), len(self._graphs)))

@@ -92,7 +297,7 @@ class Treelet(GraphKernel):
G_canonkeys = canonkeys_toshare
do_fun = self._wrapper_kernel_do
parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=init_worker,
glbv=(canonkeys,), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(canonkeys,), n_jobs=self.n_jobs, verbose=self.verbose)

return gram_matrix

@@ -104,13 +309,13 @@ class Treelet(GraphKernel):
# time, but this may cost a lot of memory for large dataset.
canonkeys_1 = self._get_canonkeys(g1)
canonkeys_list = []
iterator = get_iters(g_list, desc='getting canonkeys', file=sys.stdout, verbose=(self._verbose >= 2))
iterator = get_iters(g_list, desc='getting canonkeys', file=sys.stdout, verbose=(self.verbose >= 2))
for g in iterator:
canonkeys_list.append(self._get_canonkeys(g))

# compute kernel list.
kernel_list = [None] * len(g_list)
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self._verbose >= 2))
iterator = get_iters(range(len(g_list)), desc='Computing kernels', file=sys.stdout, length=len(g_list), verbose=(self.verbose >= 2))
for i in iterator:
kernel = self._kernel_do(canonkeys_1, canonkeys_list[i])
kernel_list[i] = kernel
@@ -125,16 +330,16 @@ class Treelet(GraphKernel):
# time, but this may cost a lot of memory for large dataset.
canonkeys_1 = self._get_canonkeys(g1)
canonkeys_list = [[] for _ in range(len(g_list))]
pool = Pool(self._n_jobs)
pool = Pool(self.n_jobs)
itr = zip(g_list, range(0, len(g_list)))
if len(g_list) < 100 * self._n_jobs:
chunksize = int(len(g_list) / self._n_jobs) + 1
if len(g_list) < 100 * self.n_jobs:
chunksize = int(len(g_list) / self.n_jobs) + 1
else:
chunksize = 100
get_fun = self._wrapper_get_canonkeys
iterator = get_iters(pool.imap_unordered(get_fun, itr, chunksize),
desc='getting canonkeys', file=sys.stdout,
length=len(g_list), verbose=(self._verbose >= 2))
length=len(g_list), verbose=(self.verbose >= 2))
for i, ck in iterator:
canonkeys_list[i] = ck
pool.close()
@@ -154,7 +359,7 @@ class Treelet(GraphKernel):
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=init_worker, glbv=(canonkeys_1, canonkeys_list), method='imap_unordered',
n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)

return kernel_list

@@ -187,7 +392,7 @@ class Treelet(GraphKernel):
keys = set(canonkey1.keys()) & set(canonkey2.keys()) # find same canonical keys in both graphs
vector1 = np.array([(canonkey1[key] if (key in canonkey1.keys()) else 0) for key in keys])
vector2 = np.array([(canonkey2[key] if (key in canonkey2.keys()) else 0) for key in keys])
kernel = self._sub_kernel(vector1, vector2)
kernel = self.sub_kernel(vector1, vector2)
return kernel


@@ -223,7 +428,7 @@ class Treelet(GraphKernel):
patterns['0'] = list(G.nodes())
canonkey['0'] = nx.number_of_nodes(G)
for i in range(1, 6): # for i in range(1, 6):
patterns[str(i)] = find_all_paths(G, i, self._ds_infos['directed'])
patterns[str(i)] = find_all_paths(G, i, self.ds_infos['directed'])
canonkey[str(i)] = len(patterns[str(i)])

# n-star patterns
@@ -317,11 +522,11 @@ class Treelet(GraphKernel):
### pattern obtained in the structural analysis section above, which is a
### string corresponding to a unique treelet. A dictionary is built to keep
### track of the amount of every treelet.
if len(self._node_labels) > 0 or len(self._edge_labels) > 0:
if len(self.node_labels) > 0 or len(self.edge_labels) > 0:
canonkey_l = {} # canonical key, a dictionary which keeps track of amount of every treelet.

# linear patterns
canonkey_t = Counter(get_mlti_dim_node_attrs(G, self._node_labels))
canonkey_t = Counter(get_mlti_dim_node_attrs(G, self.node_labels))
for key in canonkey_t:
canonkey_l[('0', key)] = canonkey_t[key]

@@ -330,9 +535,9 @@ class Treelet(GraphKernel):
for pattern in patterns[str(i)]:
canonlist = []
for idx, node in enumerate(pattern[:-1]):
canonlist.append(tuple(G.nodes[node][nl] for nl in self._node_labels))
canonlist.append(tuple(G[node][pattern[idx+1]][el] for el in self._edge_labels))
canonlist.append(tuple(G.nodes[pattern[-1]][nl] for nl in self._node_labels))
canonlist.append(tuple(G.nodes[node][nl] for nl in self.node_labels))
canonlist.append(tuple(G[node][pattern[idx+1]][el] for el in self.edge_labels))
canonlist.append(tuple(G.nodes[pattern[-1]][nl] for nl in self.node_labels))
canonkey_t = canonlist if canonlist < canonlist[::-1] else canonlist[::-1]
treelet.append(tuple([str(i)] + canonkey_t))
canonkey_l.update(Counter(treelet))
@@ -343,13 +548,13 @@ class Treelet(GraphKernel):
for pattern in patterns[str(i) + 'star']:
canonlist = []
for leaf in pattern[1:]:
nlabels = tuple(G.nodes[leaf][nl] for nl in self._node_labels)
elabels = tuple(G[leaf][pattern[0]][el] for el in self._edge_labels)
nlabels = tuple(G.nodes[leaf][nl] for nl in self.node_labels)
elabels = tuple(G[leaf][pattern[0]][el] for el in self.edge_labels)
canonlist.append(tuple((nlabels, elabels)))
canonlist.sort()
canonlist = list(chain.from_iterable(canonlist))
canonkey_t = tuple(['d' if i == 5 else str(i * 2)] +
[tuple(G.nodes[pattern[0]][nl] for nl in self._node_labels)]
[tuple(G.nodes[pattern[0]][nl] for nl in self.node_labels)]
+ canonlist)
treelet.append(canonkey_t)
canonkey_l.update(Counter(treelet))
@@ -359,17 +564,17 @@ class Treelet(GraphKernel):
for pattern in patterns['7']:
canonlist = []
for leaf in pattern[1:3]:
nlabels = tuple(G.nodes[leaf][nl] for nl in self._node_labels)
elabels = tuple(G[leaf][pattern[0]][el] for el in self._edge_labels)
nlabels = tuple(G.nodes[leaf][nl] for nl in self.node_labels)
elabels = tuple(G[leaf][pattern[0]][el] for el in self.edge_labels)
canonlist.append(tuple((nlabels, elabels)))
canonlist.sort()
canonlist = list(chain.from_iterable(canonlist))
canonkey_t = tuple(['7']
+ [tuple(G.nodes[pattern[0]][nl] for nl in self._node_labels)] + canonlist
+ [tuple(G.nodes[pattern[3]][nl] for nl in self._node_labels)]
+ [tuple(G[pattern[3]][pattern[0]][el] for el in self._edge_labels)]
+ [tuple(G.nodes[pattern[4]][nl] for nl in self._node_labels)]
+ [tuple(G[pattern[4]][pattern[3]][el] for el in self._edge_labels)])
+ [tuple(G.nodes[pattern[0]][nl] for nl in self.node_labels)] + canonlist
+ [tuple(G.nodes[pattern[3]][nl] for nl in self.node_labels)]
+ [tuple(G[pattern[3]][pattern[0]][el] for el in self.edge_labels)]
+ [tuple(G.nodes[pattern[4]][nl] for nl in self.node_labels)]
+ [tuple(G[pattern[4]][pattern[3]][el] for el in self.edge_labels)])
treelet.append(canonkey_t)
canonkey_l.update(Counter(treelet))

@@ -378,38 +583,38 @@ class Treelet(GraphKernel):
for pattern in patterns['11']:
canonlist = []
for leaf in pattern[1:4]:
nlabels = tuple(G.nodes[leaf][nl] for nl in self._node_labels)
elabels = tuple(G[leaf][pattern[0]][el] for el in self._edge_labels)
nlabels = tuple(G.nodes[leaf][nl] for nl in self.node_labels)
elabels = tuple(G[leaf][pattern[0]][el] for el in self.edge_labels)
canonlist.append(tuple((nlabels, elabels)))
canonlist.sort()
canonlist = list(chain.from_iterable(canonlist))
canonkey_t = tuple(['b']
+ [tuple(G.nodes[pattern[0]][nl] for nl in self._node_labels)] + canonlist
+ [tuple(G.nodes[pattern[4]][nl] for nl in self._node_labels)]
+ [tuple(G[pattern[4]][pattern[0]][el] for el in self._edge_labels)]
+ [tuple(G.nodes[pattern[5]][nl] for nl in self._node_labels)]
+ [tuple(G[pattern[5]][pattern[4]][el] for el in self._edge_labels)])
+ [tuple(G.nodes[pattern[0]][nl] for nl in self.node_labels)] + canonlist
+ [tuple(G.nodes[pattern[4]][nl] for nl in self.node_labels)]
+ [tuple(G[pattern[4]][pattern[0]][el] for el in self.edge_labels)]
+ [tuple(G.nodes[pattern[5]][nl] for nl in self.node_labels)]
+ [tuple(G[pattern[5]][pattern[4]][el] for el in self.edge_labels)])
treelet.append(canonkey_t)
canonkey_l.update(Counter(treelet))

# pattern 10
treelet = []
for pattern in patterns['10']:
canonkey4 = [tuple(G.nodes[pattern[5]][nl] for nl in self._node_labels),
tuple(G[pattern[5]][pattern[4]][el] for el in self._edge_labels)]
canonkey4 = [tuple(G.nodes[pattern[5]][nl] for nl in self.node_labels),
tuple(G[pattern[5]][pattern[4]][el] for el in self.edge_labels)]
canonlist = []
for leaf in pattern[1:3]:
nlabels = tuple(G.nodes[leaf][nl] for nl in self._node_labels)
elabels = tuple(G[leaf][pattern[0]][el] for el in self._edge_labels)
nlabels = tuple(G.nodes[leaf][nl] for nl in self.node_labels)
elabels = tuple(G[leaf][pattern[0]][el] for el in self.edge_labels)
canonlist.append(tuple((nlabels, elabels)))
canonlist.sort()
canonkey0 = list(chain.from_iterable(canonlist))
canonkey_t = tuple(['a']
+ [tuple(G.nodes[pattern[3]][nl] for nl in self._node_labels)]
+ [tuple(G.nodes[pattern[4]][nl] for nl in self._node_labels)]
+ [tuple(G[pattern[4]][pattern[3]][el] for el in self._edge_labels)]
+ [tuple(G.nodes[pattern[0]][nl] for nl in self._node_labels)]
+ [tuple(G[pattern[0]][pattern[3]][el] for el in self._edge_labels)]
+ [tuple(G.nodes[pattern[3]][nl] for nl in self.node_labels)]
+ [tuple(G.nodes[pattern[4]][nl] for nl in self.node_labels)]
+ [tuple(G[pattern[4]][pattern[3]][el] for el in self.edge_labels)]
+ [tuple(G.nodes[pattern[0]][nl] for nl in self.node_labels)]
+ [tuple(G[pattern[0]][pattern[3]][el] for el in self.edge_labels)]
+ canonkey4 + canonkey0)
treelet.append(canonkey_t)
canonkey_l.update(Counter(treelet))
@@ -419,15 +624,15 @@ class Treelet(GraphKernel):
for pattern in patterns['12']:
canonlist0 = []
for leaf in pattern[1:3]:
nlabels = tuple(G.nodes[leaf][nl] for nl in self._node_labels)
elabels = tuple(G[leaf][pattern[0]][el] for el in self._edge_labels)
nlabels = tuple(G.nodes[leaf][nl] for nl in self.node_labels)
elabels = tuple(G[leaf][pattern[0]][el] for el in self.edge_labels)
canonlist0.append(tuple((nlabels, elabels)))
canonlist0.sort()
canonlist0 = list(chain.from_iterable(canonlist0))
canonlist3 = []
for leaf in pattern[4:6]:
nlabels = tuple(G.nodes[leaf][nl] for nl in self._node_labels)
elabels = tuple(G[leaf][pattern[3]][el] for el in self._edge_labels)
nlabels = tuple(G.nodes[leaf][nl] for nl in self.node_labels)
elabels = tuple(G[leaf][pattern[3]][el] for el in self.edge_labels)
canonlist3.append(tuple((nlabels, elabels)))
canonlist3.sort()
canonlist3 = list(chain.from_iterable(canonlist3))
@@ -435,14 +640,14 @@ class Treelet(GraphKernel):
# 2 possible key can be generated from 2 nodes with extended label 3,
# select the one with lower lexicographic order.
canonkey_t1 = tuple(['c']
+ [tuple(G.nodes[pattern[0]][nl] for nl in self._node_labels)] + canonlist0
+ [tuple(G.nodes[pattern[3]][nl] for nl in self._node_labels)]
+ [tuple(G[pattern[3]][pattern[0]][el] for el in self._edge_labels)]
+ [tuple(G.nodes[pattern[0]][nl] for nl in self.node_labels)] + canonlist0
+ [tuple(G.nodes[pattern[3]][nl] for nl in self.node_labels)]
+ [tuple(G[pattern[3]][pattern[0]][el] for el in self.edge_labels)]
+ canonlist3)
canonkey_t2 = tuple(['c']
+ [tuple(G.nodes[pattern[3]][nl] for nl in self._node_labels)] + canonlist3
+ [tuple(G.nodes[pattern[0]][nl] for nl in self._node_labels)]
+ [tuple(G[pattern[0]][pattern[3]][el] for el in self._edge_labels)]
+ [tuple(G.nodes[pattern[3]][nl] for nl in self.node_labels)] + canonlist3
+ [tuple(G.nodes[pattern[0]][nl] for nl in self.node_labels)]
+ [tuple(G[pattern[0]][pattern[3]][el] for el in self.edge_labels)]
+ canonlist0)
treelet.append(canonkey_t1 if canonkey_t1 < canonkey_t2 else canonkey_t2)
canonkey_l.update(Counter(treelet))
@@ -450,24 +655,24 @@ class Treelet(GraphKernel):
# pattern 9
treelet = []
for pattern in patterns['9']:
canonkey2 = [tuple(G.nodes[pattern[4]][nl] for nl in self._node_labels),
tuple(G[pattern[4]][pattern[2]][el] for el in self._edge_labels)]
canonkey3 = [tuple(G.nodes[pattern[5]][nl] for nl in self._node_labels),
tuple(G[pattern[5]][pattern[3]][el] for el in self._edge_labels)]
prekey2 = [tuple(G.nodes[pattern[2]][nl] for nl in self._node_labels),
tuple(G[pattern[2]][pattern[0]][el] for el in self._edge_labels)]
prekey3 = [tuple(G.nodes[pattern[3]][nl] for nl in self._node_labels),
tuple(G[pattern[3]][pattern[0]][el] for el in self._edge_labels)]
canonkey2 = [tuple(G.nodes[pattern[4]][nl] for nl in self.node_labels),
tuple(G[pattern[4]][pattern[2]][el] for el in self.edge_labels)]
canonkey3 = [tuple(G.nodes[pattern[5]][nl] for nl in self.node_labels),
tuple(G[pattern[5]][pattern[3]][el] for el in self.edge_labels)]
prekey2 = [tuple(G.nodes[pattern[2]][nl] for nl in self.node_labels),
tuple(G[pattern[2]][pattern[0]][el] for el in self.edge_labels)]
prekey3 = [tuple(G.nodes[pattern[3]][nl] for nl in self.node_labels),
tuple(G[pattern[3]][pattern[0]][el] for el in self.edge_labels)]
if prekey2 + canonkey2 < prekey3 + canonkey3:
canonkey_t = [tuple(G.nodes[pattern[1]][nl] for nl in self._node_labels)] \
+ [tuple(G[pattern[1]][pattern[0]][el] for el in self._edge_labels)] \
canonkey_t = [tuple(G.nodes[pattern[1]][nl] for nl in self.node_labels)] \
+ [tuple(G[pattern[1]][pattern[0]][el] for el in self.edge_labels)] \
+ prekey2 + prekey3 + canonkey2 + canonkey3
else:
canonkey_t = [tuple(G.nodes[pattern[1]][nl] for nl in self._node_labels)] \
+ [tuple(G[pattern[1]][pattern[0]][el] for el in self._edge_labels)] \
canonkey_t = [tuple(G.nodes[pattern[1]][nl] for nl in self.node_labels)] \
+ [tuple(G[pattern[1]][pattern[0]][el] for el in self.edge_labels)] \
+ prekey3 + prekey2 + canonkey3 + canonkey2
treelet.append(tuple(['9']
+ [tuple(G.nodes[pattern[0]][nl] for nl in self._node_labels)]
+ [tuple(G.nodes[pattern[0]][nl] for nl in self.node_labels)]
+ canonkey_t))
canonkey_l.update(Counter(treelet))

@@ -482,12 +687,33 @@ class Treelet(GraphKernel):
return i, self._get_canonkeys(g)


def _add_dummy_labels(self, Gn):
if len(self._node_labels) == 0 or (len(self._node_labels) == 1 and self._node_labels[0] == SpecialLabel.DUMMY):
for i in range(len(Gn)):
nx.set_node_attributes(Gn[i], '0', SpecialLabel.DUMMY)
self._node_labels = [SpecialLabel.DUMMY]
if len(self._edge_labels) == 0 or (len(self._edge_labels) == 1 and self._edge_labels[0] == SpecialLabel.DUMMY):
for i in range(len(Gn)):
nx.set_edge_attributes(Gn[i], '0', SpecialLabel.DUMMY)
self._edge_labels = [SpecialLabel.DUMMY]
def _add_dummy_labels(self, Gn=None):
def _add_dummy(Gn):
if len(self.node_labels) == 0 or (len(self.node_labels) == 1 and self.node_labels[0] == SpecialLabel.DUMMY):
for i in range(len(Gn)):
nx.set_node_attributes(Gn[i], '0', SpecialLabel.DUMMY)
self.node_labels = [SpecialLabel.DUMMY]
if len(self.edge_labels) == 0 or (len(self.edge_labels) == 1 and self.edge_labels[0] == SpecialLabel.DUMMY):
for i in range(len(Gn)):
nx.set_edge_attributes(Gn[i], '0', SpecialLabel.DUMMY)
self.edge_labels = [SpecialLabel.DUMMY]

if Gn is None or Gn is self._graphs:
# Add dummy labels for the copy of self._graphs.
try:
check_is_fitted(self, ['_dummy_labels_considered'])
if not self._dummy_labels_considered:
Gn = self._graphs # @todo: ?[g.copy() for g in self._graphs]
_add_dummy(Gn)
self._graphs = Gn
self._dummy_labels_considered = True
except NotFittedError:
Gn = self._graphs # @todo: ?[g.copy() for g in self._graphs]
_add_dummy(Gn)
self._graphs = Gn
self._dummy_labels_considered = True

else:
# Add dummy labels for the input.
_add_dummy(Gn)


+ 9
- 9
gklearn/kernels/weisfeiler_lehman.py View File

@@ -33,7 +33,7 @@ class WeisfeilerLehman(GraphKernel): # @todo: sp, edge user kernel.


def _compute_gm_series(self):
# if self._verbose >= 2:
# if self.verbose >= 2:
# import warnings
# warnings.warn('A part of the computation is parallelized.')

@@ -74,17 +74,17 @@ class WeisfeilerLehman(GraphKernel): # @todo: sp, edge user kernel.
G_gn = gn_toshare
do_fun = self._wrapper_pairwise
parallel_gm(do_fun, gram_matrix, self._graphs, init_worker=init_worker,
glbv=(self._graphs,), n_jobs=self._n_jobs, verbose=self._verbose)
glbv=(self._graphs,), n_jobs=self.n_jobs, verbose=self.verbose)
return gram_matrix
else:
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('This base kernel is not parallelized. The serial computation is used instead.')
return self._compute_gm_series()


def _compute_kernel_list_series(self, g1, g_list): # @todo: this should be better.
# if self._verbose >= 2:
# if self.verbose >= 2:
# import warnings
# warnings.warn('A part of the computation is parallelized.')

@@ -126,10 +126,10 @@ class WeisfeilerLehman(GraphKernel): # @todo: sp, edge user kernel.
len_itr = len(g_list)
parallel_me(do_fun, func_assign, kernel_list, itr, len_itr=len_itr,
init_worker=init_worker, glbv=(g1, g_list), method='imap_unordered',
n_jobs=self._n_jobs, itr_desc='Computing kernels', verbose=self._verbose)
n_jobs=self.n_jobs, itr_desc='Computing kernels', verbose=self.verbose)
return kernel_list
else:
if self._verbose >= 2:
if self.verbose >= 2:
import warnings
warnings.warn('This base kernel is not parallelized. The serial computation is used instead.')
return self._compute_kernel_list_series(g1, g_list)
@@ -332,15 +332,15 @@ class WeisfeilerLehman(GraphKernel): # @todo: sp, edge user kernel.
def _compute_gram_itr(self, gram_matrix, all_num_of_each_label):
"""Compute Gram matrix using the base kernel.
"""
# if self._parallel == 'imap_unordered':
# if self.parallel == 'imap_unordered':
# # compute kernels.
# def init_worker(alllabels_toshare):
# global G_alllabels
# G_alllabels = alllabels_toshare
# do_partial = partial(self._wrapper_compute_subtree_kernel, gram_matrix)
# parallel_gm(do_partial, gram_matrix, Gn, init_worker=init_worker,
# glbv=(all_num_of_each_label,), n_jobs=self._n_jobs, verbose=self._verbose)
# elif self._parallel is None:
# glbv=(all_num_of_each_label,), n_jobs=self.n_jobs, verbose=self.verbose)
# elif self.parallel is None:
for i in range(len(gram_matrix)):
for j in range(i, len(gram_matrix)):
gram_matrix[i][j] = self._compute_subtree_kernel(all_num_of_each_label[i],


Write Preview
Loading…
Cancel
Save