|
1
|
|
|
"""!
|
|
2
|
|
|
|
|
3
|
|
|
@brief Cluster analysis algorithm: Hierarchical Sync (HSyncNet)
|
|
4
|
|
|
@details Based on article description:
|
|
5
|
|
|
- J.Shao, X.He, C.Bohm, Q.Yang, C.Plant. Synchronization-Inspired Partitioning and Hierarchical Clustering. 2013.
|
|
6
|
|
|
|
|
7
|
|
|
@authors Andrei Novikov ([email protected])
|
|
8
|
|
|
@date 2014-2016
|
|
9
|
|
|
@copyright GNU Public License
|
|
10
|
|
|
|
|
11
|
|
|
@cond GNU_PUBLIC_LICENSE
|
|
12
|
|
|
PyClustering is free software: you can redistribute it and/or modify
|
|
13
|
|
|
it under the terms of the GNU General Public License as published by
|
|
14
|
|
|
the Free Software Foundation, either version 3 of the License, or
|
|
15
|
|
|
(at your option) any later version.
|
|
16
|
|
|
|
|
17
|
|
|
PyClustering is distributed in the hope that it will be useful,
|
|
18
|
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
19
|
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
20
|
|
|
GNU General Public License for more details.
|
|
21
|
|
|
|
|
22
|
|
|
You should have received a copy of the GNU General Public License
|
|
23
|
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
24
|
|
|
@endcond
|
|
25
|
|
|
|
|
26
|
|
|
"""
|
|
27
|
|
|
|
|
28
|
|
|
import pyclustering.core.wrapper as wrapper;
|
|
29
|
|
|
|
|
30
|
|
|
from pyclustering.nnet import initial_type, solve_type;
|
|
31
|
|
|
|
|
32
|
|
|
from pyclustering.cluster.syncnet import syncnet, syncnet_analyser;
|
|
33
|
|
|
from pyclustering.utils import average_neighbor_distance;
|
|
34
|
|
|
|
|
35
|
|
|
class hsyncnet(syncnet):
|
|
36
|
|
|
"""!
|
|
37
|
|
|
@brief Class represents clustering algorithm HSyncNet. HSyncNet is bio-inspired algorithm that is based on oscillatory network that uses modified Kuramoto model.
|
|
38
|
|
|
|
|
39
|
|
|
Example:
|
|
40
|
|
|
@code
|
|
41
|
|
|
# read list of points for cluster analysis
|
|
42
|
|
|
sample = read_sample(file);
|
|
43
|
|
|
|
|
44
|
|
|
# create network for allocation three clusters using CCORE (C++ implementation)
|
|
45
|
|
|
network = hsyncnet(sample, 3, ccore = True);
|
|
46
|
|
|
|
|
47
|
|
|
# run cluster analysis and output dynamic of the network
|
|
48
|
|
|
(time, dynamic) = network.process(0.995, collect_dynamic = True);
|
|
49
|
|
|
|
|
50
|
|
|
# get allocated clusters
|
|
51
|
|
|
clusters = network.get_clusters();
|
|
52
|
|
|
|
|
53
|
|
|
# show output dynamic of the network
|
|
54
|
|
|
draw_dynamics(time, dynamic);
|
|
55
|
|
|
@endcode
|
|
56
|
|
|
"""
|
|
57
|
|
|
|
|
58
|
|
|
def __init__(self, source_data, number_clusters, osc_initial_phases = initial_type.RANDOM_GAUSSIAN, initial_neighbors = 3, increase_persent = 0.15, ccore = False):
|
|
59
|
|
|
"""!
|
|
60
|
|
|
@brief Costructor of the oscillatory network hSyncNet for cluster analysis.
|
|
61
|
|
|
|
|
62
|
|
|
@param[in] source_data (list): Input data set defines structure of the network.
|
|
63
|
|
|
@param[in] number_clusters (uint): Number of clusters that should be allocated.
|
|
64
|
|
|
@param[in] osc_initial_phases (initial_type): Type of initialization of initial values of phases of oscillators.
|
|
65
|
|
|
@param[in] initial_neighbors (uint): Defines initial radius connectivity by calculation average distance to connect specify number of oscillators.
|
|
66
|
|
|
@param[in] increase_persent (double): Percent of increasing of radius connectivity on each step (input values in range (0.0; 1.0) correspond to (0%; 100%)).
|
|
67
|
|
|
@param[in] ccore (bool): If True than DLL CCORE (C++ solution) will be used for solving.
|
|
68
|
|
|
|
|
69
|
|
|
"""
|
|
70
|
|
|
|
|
71
|
|
|
self.__ccore_network_pointer = None;
|
|
72
|
|
|
|
|
73
|
|
|
if (initial_neighbors >= len(source_data)):
|
|
74
|
|
|
initial_neighbors = len(source_data) - 1;
|
|
75
|
|
|
|
|
76
|
|
|
if (ccore is True):
|
|
77
|
|
|
self.__ccore_network_pointer = wrapper.hsyncnet_create_network(source_data, number_clusters, osc_initial_phases, initial_neighbors, increase_persent);
|
|
78
|
|
|
else:
|
|
79
|
|
|
super().__init__(source_data, 0, initial_phases = osc_initial_phases);
|
|
80
|
|
|
|
|
81
|
|
|
self.__initial_neighbors = initial_neighbors;
|
|
82
|
|
|
self.__increase_persent = increase_persent;
|
|
83
|
|
|
self._number_clusters = number_clusters;
|
|
84
|
|
|
|
|
85
|
|
|
|
|
86
|
|
|
def __del__(self):
|
|
87
|
|
|
"""!
|
|
88
|
|
|
@brief Destructor of oscillatory network HSyncNet.
|
|
89
|
|
|
|
|
90
|
|
|
"""
|
|
91
|
|
|
|
|
92
|
|
|
if (self.__ccore_network_pointer is not None):
|
|
93
|
|
|
wrapper.hsyncnet_destroy_network(self.__ccore_network_pointer);
|
|
94
|
|
|
self.__ccore_network_pointer = None;
|
|
95
|
|
|
|
|
96
|
|
|
|
|
97
|
|
|
def process(self, order = 0.998, solution = solve_type.FAST, collect_dynamic = False):
|
|
98
|
|
|
"""!
|
|
99
|
|
|
@brief Performs clustering of input data set in line with input parameters.
|
|
100
|
|
|
|
|
101
|
|
|
@param[in] order (double): Level of local synchronization between oscillator that defines end of synchronization process, range [0..1].
|
|
102
|
|
|
@param[in] solution (solve_type) Type of solving differential equation.
|
|
103
|
|
|
@param[in] collect_dynamic (bool): If True - returns whole history of process synchronization otherwise - only final state (when process of clustering is over).
|
|
104
|
|
|
|
|
105
|
|
|
@return (tuple) Returns dynamic of the network as tuple of lists on each iteration (time, oscillator_phases) that depends on collect_dynamic parameter.
|
|
106
|
|
|
|
|
107
|
|
|
@see get_clusters()
|
|
108
|
|
|
|
|
109
|
|
|
"""
|
|
110
|
|
|
|
|
111
|
|
|
if (self.__ccore_network_pointer is not None):
|
|
112
|
|
|
analyser = wrapper.hsyncnet_process(self.__ccore_network_pointer, order, solution, collect_dynamic);
|
|
113
|
|
|
return syncnet_analyser(None, None, analyser);
|
|
114
|
|
|
|
|
115
|
|
|
number_neighbors = self.__initial_neighbors;
|
|
116
|
|
|
current_number_clusters = float('inf');
|
|
117
|
|
|
|
|
118
|
|
|
dyn_phase = [];
|
|
119
|
|
|
dyn_time = [];
|
|
120
|
|
|
|
|
121
|
|
|
radius = average_neighbor_distance(self._osc_loc, number_neighbors);
|
|
122
|
|
|
|
|
123
|
|
|
increase_step = int(len(self._osc_loc) * self.__increase_persent);
|
|
124
|
|
|
if (increase_step < 1):
|
|
125
|
|
|
increase_step = 1;
|
|
126
|
|
|
|
|
127
|
|
|
|
|
128
|
|
|
analyser = None;
|
|
129
|
|
|
while(current_number_clusters > self._number_clusters):
|
|
130
|
|
|
self._create_connections(radius);
|
|
131
|
|
|
|
|
132
|
|
|
analyser = self.simulate_dynamic(order, solution, collect_dynamic);
|
|
133
|
|
|
if (collect_dynamic == True):
|
|
134
|
|
|
dyn_phase += analyser.output;
|
|
135
|
|
|
|
|
136
|
|
|
if (len(dyn_time) > 0):
|
|
137
|
|
|
point_time_last = dyn_time[len(dyn_time) - 1];
|
|
138
|
|
|
dyn_time += [time_point + point_time_last for time_point in analyser.time];
|
|
139
|
|
|
else:
|
|
140
|
|
|
dyn_time += analyser.time;
|
|
141
|
|
|
|
|
142
|
|
|
clusters = analyser.allocate_sync_ensembles(0.05);
|
|
143
|
|
|
|
|
144
|
|
|
# Get current number of allocated clusters
|
|
145
|
|
|
current_number_clusters = len(clusters);
|
|
146
|
|
|
|
|
147
|
|
|
# Increase number of neighbors that should be used
|
|
148
|
|
|
number_neighbors += increase_step;
|
|
149
|
|
|
|
|
150
|
|
|
# Update connectivity radius and check if average function can be used anymore
|
|
151
|
|
|
if (number_neighbors >= len(self._osc_loc)):
|
|
152
|
|
|
radius = radius * self.__increase_persent + radius;
|
|
153
|
|
|
else:
|
|
154
|
|
|
radius = average_neighbor_distance(self._osc_loc, number_neighbors);
|
|
155
|
|
|
|
|
156
|
|
|
if (collect_dynamic != True):
|
|
157
|
|
|
dyn_phase = analyser.output;
|
|
158
|
|
|
dyn_time = analyser.time;
|
|
159
|
|
|
|
|
160
|
|
|
return syncnet_analyser(dyn_phase, dyn_time, None);
|
|
161
|
|
|
|
annoviko /
pyclustering
