notebook.community

Edit and run

Aim

Encode a few simple stick figures that the layout algorithm is likely to display correctly.

Generate a grammar and observe the derived samples.





In [1]:



    


from logging import getLogger 
from eden.util import configure_logging
configure_logging(getLogger(),verbosity=1)



    











In [2]:



    


%matplotlib inline
#draw stick figures
import networkx as nx



    











In [3]:



    


G=nx.Graph()
G.add_node(0, label='head')
G.add_node(1, label='neck')
G.add_node(11, label='b')
G.add_node(12, label='b')
G.add_node(13, label='b')
G.add_node(22, label='b')
G.add_node(23, label='b')
G.add_node(32, label='b')
G.add_node(33, label='b')
G.add_node(5, label='arm')
G.add_node(6, label='arm')
G.add_node(55, label='arm')
G.add_node(56, label='arm')
G.add_node(7, label='leg')
G.add_node(8, label='leg')
G.add_node(57, label='leg')
G.add_node(58, label='leg')
G.add_node(9, label='pelvis')
G.add_edge(0,1, label='.')
G.add_edge(1,22, label='.')
G.add_edge(1,23, label='.')
G.add_edge(9,11, label='.')
G.add_edge(11,12, label='.')
G.add_edge(11,13, label='.')
G.add_edge(12,32, label='.')
G.add_edge(13,33, label='.')
G.add_edge(22,32, label='.')
G.add_edge(23,33, label='.')
G.add_edge(9,7, label='.')
G.add_edge(9,8, label='.')
G.add_edge(22,5, label='.')
G.add_edge(23,6, label='.')
G.add_edge(5,55, label='.')
G.add_edge(6,56, label='.')
G.add_edge(7,57, label='.')
G.add_edge(8,58, label='.')


from eden.util import display
display.draw_graph(G, size=5, node_size=500, prog='circo')

g1 = G.copy()



    


















    
























In [4]:



    


G=nx.Graph()
G.add_node(0, label='head2')
G.add_node(1, label='neck')
G.add_node(11, label='b')
G.add_node(12, label='b')
G.add_node(13, label='b')
G.add_node(22, label='b')
G.add_node(23, label='b')
G.add_node(32, label='b')
G.add_node(33, label='b')
G.add_node(42, label='b')
G.add_node(43, label='b')
G.add_node(5, label='arm')
G.add_node(6, label='arm')
G.add_node(55, label='arm2')
G.add_node(56, label='arm2')
G.add_node(7, label='leg')
G.add_node(8, label='leg')
G.add_node(57, label='leg2')
G.add_node(58, label='leg2')
G.add_node(9, label='pelvis')
G.add_edge(0,1, label='.')
G.add_edge(1,22, label='.')
G.add_edge(1,23, label='.')
G.add_edge(9,11, label='.')
G.add_edge(11,42, label='.')
G.add_edge(11,43, label='.')
G.add_edge(42,12, label='.')
G.add_edge(43,13, label='.')
G.add_edge(12,32, label='.')
G.add_edge(13,33, label='.')
G.add_edge(22,32, label='.')
G.add_edge(23,33, label='.')
G.add_edge(9,7, label='.')
G.add_edge(9,8, label='.')
G.add_edge(22,5, label='.')
G.add_edge(23,6, label='.')
G.add_edge(5,55, label='.')
G.add_edge(6,56, label='.')
G.add_edge(7,57, label='.')
G.add_edge(8,58, label='.')

from eden.util import display
display.draw_graph(G, size=5, node_size=500, prog='circo')

g2 = G.copy()



    


















    
























In [35]:



    


from graphlearn.graphlearn import GraphLearnSampler
from eden.converter.graph.gspan import gspan_to_eden
from eden.graph import Vectorizer
import itertools
gr = [g2,g1,g1,g1]
gr = [g1,g2,g2,g2]



sampler=GraphLearnSampler(radius_list=[0,1],thickness_list=[1,2],
                          min_cip_count=1, min_interface_count=1,
                         vectorizer=Vectorizer(5))
sampler.fit(gr)



    











In [36]:



    


from graphlearn.utils.draw import draw_grammar
draw_grammar(sampler.grammar().productions,
             n_productions=15,n_graphs_per_line=6,
             size=5, colormap='Paired', invert_colormap=False,node_border=1,vertex_alpha=0.6, edge_alpha=0.2, node_size=450)



    


















    






interface id: 892921 [4 options]










    

















    






interface id: 884950 [2 options]










    

















    






interface id: 881865 [2 options]










    

















    






interface id: 835491 [2 options]










    

















    






interface id: 790151 [2 options]










    

















    






interface id: 652364 [2 options]










    

















    






interface id: 644272 [2 options]










    

















    






interface id: 553050 [2 options]










    

















    






interface id: 527088 [2 options]










    

















    






interface id: 465702 [2 options]










    

















    






interface id: 431002 [2 options]










    

















    






interface id: 411421 [2 options]










    

















    






interface id: 340226 [2 options]










    

















    






interface id: 297786 [2 options]










    

















    






interface id: 268796 [2 options]










    
























In [37]:



    


#sample
graphs = [g2]*4
graphs = [g1]*4
n_steps=40
n_samples=6
graphs = sampler.sample(graphs,
                        n_steps=n_steps, n_samples=n_samples,
                        target_orig_cip=True,
                        probabilistic_core_choice=False,
                        score_core_choice= False,
                        max_core_size_diff=1,
                        burnin=1,
                        omit_seed=False,
                        generator_mode=False,
                        improving_threshold=0.5, 
                        accept_static_penalty=0,
                        n_jobs=-1,
                        select_cip_max_tries=200,
                        keep_duplicates=True)



    











In [38]:



    


%%time
# plot examples of sampling paths
from graphlearn.utils.draw import  graphlearn_draw, get_score_of_graph
scores=[]
for i,graph in enumerate(graphs):
    print 'Graph id: %d'%(i)
    scores.append(graph.graph['sampling_info']['score_history'])
    path_graphs = graph.graph['sampling_info']['graphs_history']
    graphlearn_draw(path_graphs,
                   n_graphs_per_line=6, size=7, 
                   colormap='Paired', invert_colormap=False,node_border=0.5, vertex_color='_labels_',
                   vertex_alpha=0.5, edge_alpha=0.2, node_size=450,
                   headlinehook=get_score_of_graph)



    


















    






Graph id: 0










    

















    






Graph id: 1










    

















    






Graph id: 2










    

















    






Graph id: 3










    

















    






CPU times: user 6.09 s, sys: 666 ms, total: 6.76 s
Wall time: 47 s

















In [39]:



    


%%time
# plot sampling path score
from itertools import islice
import numpy as np
import pylab as plt
markevery=n_steps/(n_samples-1)
step=1
num_graphs_per_plot=4
num_plots=np.ceil([len(scores)/num_graphs_per_plot])
for i in range(num_plots):
    plt.figure(figsize=(13,5))
    for j,score in enumerate(scores[i*num_graphs_per_plot:i*num_graphs_per_plot+num_graphs_per_plot]):
        data = list(islice(score,None, None, step))
        plt.plot(data, linewidth=2, label='graph %d'%(j+i*num_graphs_per_plot))
        plt.plot(data, linestyle='None',markevery=markevery, markerfacecolor='white', marker='o', markeredgewidth=2,markersize=6)
    plt.legend(loc='lower right')
    plt.grid()
    plt.xlim(-1,n_steps+1)
    plt.ylim(-0.1,1.1)
    plt.show()



    


















    

















    






CPU times: user 244 ms, sys: 8.58 ms, total: 252 ms
Wall time: 251 ms

.

Content source: fabriziocosta/GraphLearn_examples

Similar notebooks:

notebook.community | gallery | about