Python networkx.disjoint_union_all() Examples

def n_community(c_sizes, p_inter=0.01):
    graphs = [nx.gnp_random_graph(c_sizes[i], 0.7, seed=i) for i in range(len(c_sizes))]
    G = nx.disjoint_union_all(graphs)
    communities = list(nx.connected_component_subgraphs(G))
    for i in range(len(communities)):
        subG1 = communities[i]
        nodes1 = list(subG1.nodes())
        for j in range(i+1, len(communities)):
            subG2 = communities[j]
            nodes2 = list(subG2.nodes())
            has_inter_edge = False
            for n1 in nodes1:
                for n2 in nodes2:
                    if np.random.rand() < p_inter:
                        G.add_edge(n1, n2)
                        has_inter_edge = True
            if not has_inter_edge:
                G.add_edge(nodes1[0], nodes2[0])
    #print('connected comp: ', len(list(nx.connected_component_subgraphs(G))))
    return G 

def n_community(c_sizes, p_inter=0.01):
    graphs = [nx.gnp_random_graph(c_sizes[i], 0.7, seed=i) for i in range(len(c_sizes))]
    G = nx.disjoint_union_all(graphs)
    communities = list(nx.connected_component_subgraphs(G))
    for i in range(len(communities)):
        subG1 = communities[i]
        nodes1 = list(subG1.nodes())
        for j in range(i+1, len(communities)):
            subG2 = communities[j]
            nodes2 = list(subG2.nodes())
            has_inter_edge = False
            for n1 in nodes1:
                for n2 in nodes2:
                    if np.random.rand() < p_inter:
                        G.add_edge(n1, n2)
                        has_inter_edge = True
            if not has_inter_edge:
                G.add_edge(nodes1[0], nodes2[0])
    #print('connected comp: ', len(list(nx.connected_component_subgraphs(G))))
    return G 

def disjoint_cliques_test_graph(num_cliques, clique_size):
    G = nx.disjoint_union_all([nx.complete_graph(clique_size) for _ in range(num_cliques)])
    return nx.to_numpy_matrix(G) 

def disjoint_cliques_test_graph(num_cliques, clique_size):
    G = nx.disjoint_union_all([nx.complete_graph(clique_size) for _ in range(num_cliques)])
    return nx.to_numpy_matrix(G) 

def gen_2hier(num_graphs, num_clusters, n, m_range, inter_prob1, inter_prob2, feat_gen):
    ''' Each community is a BA graph.
    Args:
        inter_prob1: probability of one node connecting to any node in the other community within
            the large cluster.
        inter_prob2: probability of one node connecting to any node in the other community between
            the large cluster.
    '''
    graphs = []

    for i in range(num_graphs):
        clusters2 = []
        for j in range(len(num_clusters)):
            clusters = gen_er(range(n, n+1), 0.5, num_clusters[j], feat_gen[0])
            G = nx.disjoint_union_all(clusters)
            for u1 in range(G.number_of_nodes()):
                if np.random.rand() < inter_prob1:
                    target = np.random.choice(G.number_of_nodes() - n)
                    # move one cluster after to make sure it's not an intra-cluster edge
                    if target // n >= u1 // n:
                        target += n
                    G.add_edge(u1, target)
            clusters2.append(G)
        G = nx.disjoint_union_all(clusters2)
        cluster_sizes_cum = np.cumsum([cluster2.number_of_nodes() for cluster2 in clusters2])
        curr_cluster = 0
        for u1 in range(G.number_of_nodes()):
            if u1 >= cluster_sizes_cum[curr_cluster]:
                curr_cluster += 1
            if np.random.rand() < inter_prob2:
                target = np.random.choice(G.number_of_nodes() -
                        clusters2[curr_cluster].number_of_nodes())
                # move one cluster after to make sure it's not an intra-cluster edge
                if curr_cluster == 0 or target >= cluster_sizes_cum[curr_cluster - 1]:
                    target += cluster_sizes_cum[curr_cluster]
            G.add_edge(u1, target)
        graphs.append(G)

    return graphs 

def test_mixed_type_disjoint_union():
    G = nx.Graph()
    H = nx.MultiGraph()
    I = nx.Graph()
    U = nx.disjoint_union_all([G,H,I]) 

def test_input_output():
    l = [nx.Graph([(1,2)]),nx.Graph([(3,4)])]
    U = nx.disjoint_union_all(l)
    assert_equal(len(l),2)
    C = nx.compose_all(l)
    assert_equal(len(l),2)
    l = [nx.Graph([(1,2)]),nx.Graph([(1,2)])]
    R = nx.intersection_all(l)
    assert_equal(len(l),2) 

def test_mixed_type_disjoint_union():
    G = nx.Graph()
    H = nx.MultiGraph()
    I = nx.Graph()
    U = nx.disjoint_union_all([G,H,I]) 

def test_input_output():
    l = [nx.Graph([(1,2)]),nx.Graph([(3,4)])]
    U = nx.disjoint_union_all(l)
    assert_equal(len(l),2)
    C = nx.compose_all(l)
    assert_equal(len(l),2)
    l = [nx.Graph([(1,2)]),nx.Graph([(1,2)])]
    R = nx.intersection_all(l)
    assert_equal(len(l),2) 

def test_input_output():
    l = [nx.Graph([(1, 2)]), nx.Graph([(3, 4)])]
    U = nx.disjoint_union_all(l)
    assert_equal(len(l), 2)
    C = nx.compose_all(l)
    assert_equal(len(l), 2)
    l = [nx.Graph([(1, 2)]), nx.Graph([(1, 2)])]
    R = nx.intersection_all(l)
    assert_equal(len(l), 2) 

def test_mixed_type_disjoint_union():
    G = nx.Graph()
    H = nx.MultiGraph()
    I = nx.Graph()
    U = nx.disjoint_union_all([G, H, I]) 

def test_empty_disjoint_union():
    nx.disjoint_union_all([]) 

def windmill_graph(n, k):
    """Generate a windmill graph.
    A windmill graph is a graph of `n` cliques each of size `k` that are all
    joined at one node.
    It can be thought of as taking a disjoint union of `n` cliques of size `k`,
    selecting one point from each, and contracting all of the selected points.
    Alternatively, one could generate `n` cliques of size `k-1` and one node
    that is connected to all other nodes in the graph.

    Parameters
    ----------
    n : int
      Number of cliques
    k : int
      Size of cliques

    Returns
    -------
    G : NetworkX Graph
      windmill graph with n cliques of size k

    Raises
    ------
    NetworkXError
        If the number of cliques is less than two
        If the size of the cliques are less than two

    Examples
    --------
    >>> G = nx.windmill_graph(4, 5)

    Notes
    -----
    The node labeled `0` will be the node connected to all other nodes.
    Note that windmill graphs are usually denoted `Wd(k,n)`, so the parameters
    are in the opposite order as the parameters of this method.
    """
    if n < 2:
        msg = 'A windmill graph must have at least two cliques'
        raise nx.NetworkXError(msg)
    if k < 2:
        raise nx.NetworkXError('The cliques must have at least two nodes')

    G = nx.disjoint_union_all(itertools.chain([nx.complete_graph(k)],
                                              (nx.complete_graph(k - 1)
                                               for _ in range(n - 1))))
    G.add_edges_from((0, i) for i in range(k, G.number_of_nodes()))
    return G 

def test_union_all_and_compose_all():
    K3=nx.complete_graph(3)
    P3=nx.path_graph(3)

    G1=nx.DiGraph()
    G1.add_edge('A','B')
    G1.add_edge('A','C')
    G1.add_edge('A','D')
    G2=nx.DiGraph()
    G2.add_edge('1','2')
    G2.add_edge('1','3')
    G2.add_edge('1','4')

    G=nx.union_all([G1,G2])
    H=nx.compose_all([G1,G2])
    assert_edges_equal(G.edges(),H.edges())
    assert_false(G.has_edge('A','1'))
    assert_raises(nx.NetworkXError, nx.union, K3, P3)
    H1=nx.union_all([H,G1],rename=('H','G1'))
    assert_equal(sorted(H1.nodes()),
        ['G1A', 'G1B', 'G1C', 'G1D',
         'H1', 'H2', 'H3', 'H4', 'HA', 'HB', 'HC', 'HD'])

    H2=nx.union_all([H,G2],rename=("H",""))
    assert_equal(sorted(H2.nodes()),
        ['1', '2', '3', '4',
         'H1', 'H2', 'H3', 'H4', 'HA', 'HB', 'HC', 'HD'])

    assert_false(H1.has_edge('NB','NA'))

    G=nx.compose_all([G,G])
    assert_edges_equal(G.edges(),H.edges())

    G2=nx.union_all([G2,G2],rename=('','copy'))
    assert_equal(sorted(G2.nodes()),
        ['1', '2', '3', '4', 'copy1', 'copy2', 'copy3', 'copy4'])

    assert_equal(sorted(G2.neighbors('copy4')),[])
    assert_equal(sorted(G2.neighbors('copy1')),['copy2', 'copy3', 'copy4'])
    assert_equal(len(G),8)
    assert_equal(nx.number_of_edges(G),6)

    E=nx.disjoint_union_all([G,G])
    assert_equal(len(E),16)
    assert_equal(nx.number_of_edges(E),12)

    E=nx.disjoint_union_all([G1,G2])
    assert_equal(sorted(E.nodes()),[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11])

    G1=nx.DiGraph()
    G1.add_edge('A','B')
    G2=nx.DiGraph()
    G2.add_edge(1,2)
    G3=nx.DiGraph()
    G3.add_edge(11,22)
    G4=nx.union_all([G1,G2,G3],rename=("G1","G2","G3"))
    assert_equal(sorted(G4.nodes()),
        ['G1A', 'G1B', 'G21', 'G22',
         'G311', 'G322']) 

def windmill_graph(n, k):
    """Generate a windmill graph.
    A windmill graph is a graph of `n` cliques each of size `k` that are all
    joined at one node.
    It can be thought of as taking a disjoint union of `n` cliques of size `k`,
    selecting one point from each, and contracting all of the selected points.
    Alternatively, one could generate `n` cliques of size `k-1` and one node
    that is connected to all other nodes in the graph.

    Parameters
    ----------
    n : int
        Number of cliques
    k : int
        Size of cliques

    Returns
    -------
    G : NetworkX Graph
        windmill graph with n cliques of size k

    Raises
    ------
    NetworkXError
        If the number of cliques is less than two
        If the size of the cliques are less than two

    Examples
    --------
    >>> G = nx.windmill_graph(4, 5)

    Notes
    -----
    The node labeled `0` will be the node connected to all other nodes.
    Note that windmill graphs are usually denoted `Wd(k,n)`, so the parameters
    are in the opposite order as the parameters of this method.
    """
    if n < 2:
        msg = 'A windmill graph must have at least two cliques'
        raise nx.NetworkXError(msg)
    if k < 2:
        raise nx.NetworkXError('The cliques must have at least two nodes')

    G = nx.disjoint_union_all(itertools.chain([nx.complete_graph(k)],
                                              (nx.complete_graph(k - 1)
                                               for _ in range(n - 1))))
    G.add_edges_from((0, i) for i in range(k, G.number_of_nodes()))
    return G 

def test_union_all_and_compose_all():
    K3 = nx.complete_graph(3)
    P3 = nx.path_graph(3)

    G1 = nx.DiGraph()
    G1.add_edge('A', 'B')
    G1.add_edge('A', 'C')
    G1.add_edge('A', 'D')
    G2 = nx.DiGraph()
    G2.add_edge('1', '2')
    G2.add_edge('1', '3')
    G2.add_edge('1', '4')

    G = nx.union_all([G1, G2])
    H = nx.compose_all([G1, G2])
    assert_edges_equal(G.edges(), H.edges())
    assert_false(G.has_edge('A', '1'))
    assert_raises(nx.NetworkXError, nx.union, K3, P3)
    H1 = nx.union_all([H, G1], rename=('H', 'G1'))
    assert_equal(sorted(H1.nodes()),
                 ['G1A', 'G1B', 'G1C', 'G1D',
                  'H1', 'H2', 'H3', 'H4', 'HA', 'HB', 'HC', 'HD'])

    H2 = nx.union_all([H, G2], rename=("H", ""))
    assert_equal(sorted(H2.nodes()),
                 ['1', '2', '3', '4',
                  'H1', 'H2', 'H3', 'H4', 'HA', 'HB', 'HC', 'HD'])

    assert_false(H1.has_edge('NB', 'NA'))

    G = nx.compose_all([G, G])
    assert_edges_equal(G.edges(), H.edges())

    G2 = nx.union_all([G2, G2], rename=('', 'copy'))
    assert_equal(sorted(G2.nodes()),
                 ['1', '2', '3', '4', 'copy1', 'copy2', 'copy3', 'copy4'])

    assert_equal(sorted(G2.neighbors('copy4')), [])
    assert_equal(sorted(G2.neighbors('copy1')), ['copy2', 'copy3', 'copy4'])
    assert_equal(len(G), 8)
    assert_equal(nx.number_of_edges(G), 6)

    E = nx.disjoint_union_all([G, G])
    assert_equal(len(E), 16)
    assert_equal(nx.number_of_edges(E), 12)

    E = nx.disjoint_union_all([G1, G2])
    assert_equal(sorted(E.nodes()), [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11])

    G1 = nx.DiGraph()
    G1.add_edge('A', 'B')
    G2 = nx.DiGraph()
    G2.add_edge(1, 2)
    G3 = nx.DiGraph()
    G3.add_edge(11, 22)
    G4 = nx.union_all([G1, G2, G3], rename=("G1", "G2", "G3"))
    assert_equal(sorted(G4.nodes()),
                 ['G1A', 'G1B', 'G21', 'G22',
                  'G311', 'G322']) 

def test_union_all_and_compose_all():
    K3=nx.complete_graph(3)
    P3=nx.path_graph(3)

    G1=nx.DiGraph()
    G1.add_edge('A','B')
    G1.add_edge('A','C')
    G1.add_edge('A','D')
    G2=nx.DiGraph()
    G2.add_edge('1','2')
    G2.add_edge('1','3')
    G2.add_edge('1','4')

    G=nx.union_all([G1,G2])
    H=nx.compose_all([G1,G2])
    assert_edges_equal(G.edges(),H.edges())
    assert_false(G.has_edge('A','1'))
    assert_raises(nx.NetworkXError, nx.union, K3, P3)
    H1=nx.union_all([H,G1],rename=('H','G1'))
    assert_equal(sorted(H1.nodes()),
        ['G1A', 'G1B', 'G1C', 'G1D',
         'H1', 'H2', 'H3', 'H4', 'HA', 'HB', 'HC', 'HD'])

    H2=nx.union_all([H,G2],rename=("H",""))
    assert_equal(sorted(H2.nodes()),
        ['1', '2', '3', '4',
         'H1', 'H2', 'H3', 'H4', 'HA', 'HB', 'HC', 'HD'])

    assert_false(H1.has_edge('NB','NA'))

    G=nx.compose_all([G,G])
    assert_edges_equal(G.edges(),H.edges())

    G2=nx.union_all([G2,G2],rename=('','copy'))
    assert_equal(sorted(G2.nodes()),
        ['1', '2', '3', '4', 'copy1', 'copy2', 'copy3', 'copy4'])

    assert_equal(G2.neighbors('copy4'),[])
    assert_equal(sorted(G2.neighbors('copy1')),['copy2', 'copy3', 'copy4'])
    assert_equal(len(G),8)
    assert_equal(nx.number_of_edges(G),6)

    E=nx.disjoint_union_all([G,G])
    assert_equal(len(E),16)
    assert_equal(nx.number_of_edges(E),12)

    E=nx.disjoint_union_all([G1,G2])
    assert_equal(sorted(E.nodes()),[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11])

    G1=nx.DiGraph()
    G1.add_edge('A','B')
    G2=nx.DiGraph()
    G2.add_edge(1,2)
    G3=nx.DiGraph()
    G3.add_edge(11,22)
    G4=nx.union_all([G1,G2,G3],rename=("G1","G2","G3"))
    assert_equal(sorted(G4.nodes()),
        ['G1A', 'G1B', 'G21', 'G22',
         'G311', 'G322']) 

def enron_task(args, idx=None, writer=None):
    labels_dict = {
        "None": 5,
        "Employee": 0,
        "Vice President": 1,
        "Manager": 2,
        "Trader": 3,
        "CEO+Managing Director+Director+President": 4,
    }
    max_enron_id = 183
    if idx is None:
        G_list = []
        labels_list = []
        for i in range(10):
            net = pickle.load(
                open("data/gnn-explainer-enron/enron_slice_{}.pkl".format(i), "rb")
            )
            # net.add_nodes_from(range(max_enron_id))
            # labels=[n[1].get('role', 'None') for n in net.nodes(data=True)]
            # labels_num = [labels_dict[l] for l in labels]
            featgen_const = featgen.ConstFeatureGen(
                np.ones(args.input_dim, dtype=float)
            )
            featgen_const.gen_node_features(net)
            G_list.append(net)
            print(net.number_of_nodes())
            # labels_list.append(labels_num)

        G = nx.disjoint_union_all(G_list)
        model = models.GcnEncoderNode(
            args.input_dim,
            args.hidden_dim,
            args.output_dim,
            len(labels_dict),
            args.num_gc_layers,
            bn=args.bn,
            args=args,
        )
        labels = [n[1].get("role", "None") for n in G.nodes(data=True)]
        labels_num = [labels_dict[l] for l in labels]
        for i in range(5):
            print("Label ", i, ": ", labels_num.count(i))

        print("Total num nodes: ", len(labels_num))
        print(labels_num)

        if args.gpu:
            model = model.cuda()
        train_node_classifier(G, labels_num, model, args, writer=writer)
    else:
        print("Running Enron full task")