path: root/src/make_plots.py

import matplotlib.pyplot as plt
import numpy as np
import cascade_creation
import convex_optimization
import algorithms
import timeout
import rip_condition


def theoreticalGuarantees(graph_name, n_cascades, min_proba, max_proba,
        sparse_edges, p_init, passed_function, *args, **kwargs):

    #creating original graph
    G = cascade_creation.InfluenceGraph(max_proba = max_proba,
                                        min_proba = min_proba,
                                        sparse_edges = sparse_edges
                                        )
    G.import_from_file(graph_name)

    #creating dummy graph
    G_hat = cascade_creation.InfluenceGraph(max_proba=None)
    G_hat.add_nodes_from(G.nodes())

    #generating cascades
    A = cascade_creation.generate_cascades(G, p_init=p_init,
                                            n_cascades = n_cascades)

    #Value of restrictedEigenvalueList
    restrictedEigenvalueList = []

    for node in G.nodes():
        print(node)
        try:
            M, _ = cascade_creation.icc_matrixvector_for_node(A, node)
            M_val = np.delete(M, node, axis=1)
            parents = np.nonzero(G.mat[:,node])[0]
            restrictedEigenvalueList.append(
                    convex_optimization.restrictedEigenvalue(M, S=parents,
                        gramMatrix=True, node=node))
            print(restrictedEigenvalueList[-1])
        except timeout.TimeoutError:
            print("TimeoutError, skipping to next node")

    print(restrictedEigenvalueList)
    alpha = 1
    theoreticalGuarantee = sum(
            np.sqrt(len(parents)*np.log(G.number_of_nodes())/len(M))/(
                alpha*gamma) for gamma in restrictedEigenvalueList if gamma >=
            1000)
    print(theoreticalGuarantee)


def cascades_vs_measurements(graph_name, n_cascades, min_proba, max_proba,
        sparse_edges=False, p_init=0.05):
    """
    Compares the number of measurements within a cascade for different types of
    graphs
    """
    G = cascade_creation.InfluenceGraph(max_proba = max_proba,
                                        min_proba = min_proba,
                                        sparse_edges = sparse_edges
                                        )
    G.import_from_file(graph_name)
    A = cascade_creation.generate_cascades(G, p_init=p_init,
                                            n_cascades = n_cascades)

    L = [len(cascade) for cascade in A]
    print(L)
    print("mean: {}\nstandard deviation: {}".format(np.mean(L), np.std(L)))


def compare_greedy_and_lagrange_cs284r():
    """
    Compares the performance of the greedy algorithm on the
    lagrangian sparse recovery objective on the Facebook dataset
    for the CS284r project
    """
    G = cascade_creation.InfluenceGraph(max_proba = .8)
    G.import_from_file("../datasets/subset_facebook_SNAPnormalize.txt")
    A = cascade_creation.generate_cascades(G, p_init=.05, n_cascades=100)

    #Greedy
    G_hat = algorithms.greedy_prediction(G, A)
    algorithms.correctness_measure(G, G_hat, print_values=True)

    #Lagrange Objective
    G_hat = algorithms.recovery_l1obj_l2constraint(G, A,
            passed_function=convex_optimization.type_lasso,
            floor_cstt=.05, lbda=10)
    algorithms.correctness_measure(G, G_hat, print_values=True)


def compute_graph(graph_name, n_cascades, lbda, passed_function, min_proba,
                max_proba, sparse_edges=False, p_init=.05):
    """
    Test running time on different algorithms
    """
    G = cascade_creation.InfluenceGraph(max_proba=max_proba,
                                        min_proba=min_proba,
                                        sparse_edges=sparse_edges)
    G.import_from_file(graph_name)
    A = cascade_creation.generate_cascades(G, p_init=p_init,
            n_cascades=n_cascades)

    if passed_function==algorithms.greedy_prediction:
        G_hat = algorithms.greedy_prediction(G, A)
    else:
        G_hat = algorithms.recovery_passed_function(G, A,
                passed_function=passed_function,
                floor_cstt=.1, lbda=lbda, n_cascades=n_cascades)
    algorithms.correctness_measure(G, G_hat, print_values=True)


def plot_watts_strogatz_graph():
    """
    plot information in a pretty way
    """
    plt.clf()
    fig = plt.figure(1)
    labels = [50, 100, 500, 1000, 2000, 5000]
    x = [np.log(50), np.log(100), np.log(500),
            np.log(1000), np.log(2000), np.log(5000)]
    sparse_recov = [.25, .32, .7, .82, .89, .92]
    max_likel = [.21, .29, .67, .8, .87, .9]
    lasso = [.07, .30, .46, .65, .86, .89]
    greedy = [.09, .15, .4, .63, .82, .92]

    fig, ax = plt.subplots()

    plt.axis((np.log(45), np.log(5500), 0, 1))
    plt.xlabel("Number of Cascades")
    plt.ylabel("F1 score")
    plt.grid(color="lightgrey")
    ax.plot(x, greedy, 'ko-', color="lightseagreen", label='Greedy')
    ax.plot(x, lasso, 'ko-', color="orange", label="Lasso")
    ax.plot(x, max_likel, 'ko-', color="cornflowerblue", label="MLE")
    ax.plot(x, sparse_recov, 'ko-', color="k", label="Our Method")
    plt.legend(loc="lower right", fontsize=18)
    ax.set_xticks(x)
    ax.set_xticklabels(tuple(labels))
    plt.savefig("../paper/figures/"+"watts_strogatz.pdf")


def plot_barabasi_albert_graph():
    """
    plot information in a pretty way
    """
    plt.clf()
    fig = plt.figure(1)
    labels = [50, 100, 500, 1000, 2000, 5000]
    x = [np.log(50), np.log(100), np.log(500),
            np.log(1000), np.log(2000), np.log(5000)]
    sparse_recov = [.35, .38, .58, .69, .79, .86]
    max_likel = [.35, .38, .56, .68, .78, .85]
    lasso = [.25, .3, .55, .67, .76, .83]
    greedy = [.1, .13, .33, .47, .6, .75]

    fig, ax = plt.subplots()

    plt.axis((np.log(45), np.log(5500), 0, 1))
    plt.xlabel("Number of Cascades")
    plt.ylabel("F1 score")
    plt.grid(color="lightgrey")
    ax.plot(x, greedy, 'ko-', color="lightseagreen", label='Greedy')
    ax.plot(x, lasso, 'ko-', color="orange", label="Lasso")
    ax.plot(x, max_likel, 'ko-', color="cornflowerblue", label="MLE")
    ax.plot(x, sparse_recov, 'ko-', color="k", label="Our Method")
    plt.legend(loc="lower right", fontsize=18)
    ax.set_xticks(x)
    ax.set_xticklabels(tuple(labels))

    plt.savefig("../paper/figures/"+"barabasi_albert.pdf")



def plot_kronecker_l2norm():
    plt.clf()
    fig = plt.figure(1)
    x = [50, 150, 500, 1000, 2000]
    sparse_recov = [62, 60, 36, 28, 21]
    max_likel = [139, 101, 42, 31, 25]
    lasso = [50, 48, 33, 29, 23]

    fig, ax = plt.subplots()

    plt.subplots_adjust(bottom=.2, top=.85)
    plt.xticks(ha="right", rotation=45)

    plt.axis((50, 2000, 0, 145))
    plt.xlabel("Number of Cascades")
    plt.ylabel("l2-norm")
    plt.grid(color="lightgrey")
    ax.plot(x, lasso, 'ko-', color="orange", label="Lasso")
    ax.plot(x, max_likel, 'ko-', color="cornflowerblue", label="MLE")
    ax.plot(x, sparse_recov, 'ko-', color="k", label="Our Method")
    plt.legend(loc="upper right")
    ax.set_xticks(x)
    ax.set_xticklabels(tuple([50, 100, 500, 1000, 2000]))
    ax.set_yticklabels(tuple(['', 20, 40, 60, 80, 100, 120, 140]))
    plt.savefig("../paper/figures/"+"kronecker_l2_norm.pdf")


def plot_kronecker_l2norm_nonsparse():
    plt.clf()
    fig = plt.figure(1)
    x = [50, 150, 500, 1000, 2000]
    sparse_recov = [56, 55, 28, 21, 15]
    max_likel = [125, 80, 35, 25, 20]
    lasso = [47, 47, 27, 22, 17]

    fig, ax = plt.subplots()

    plt.subplots_adjust(bottom=.2, top=.85)
    plt.xticks(ha="right", rotation=45)

    plt.axis((50, 2000, 0, 145))
    plt.xlabel("Number of Cascades")
    plt.ylabel("l2-norm")
    plt.grid(color="lightgrey")
    ax.plot(x, lasso, 'ko-', color="orange", label="Lasso")
    ax.plot(x, max_likel, 'ko-', color="cornflowerblue", label="MLE")
    ax.plot(x, sparse_recov, 'ko-', color="k", label="Our Method")
    plt.legend(loc="upper right")
    ax.set_xticks(x)
    ax.set_xticklabels(tuple([50, 100, 500, 1000, 2000]))
    ax.set_yticklabels(tuple(['', 20, 40, 60, 80, 100, 120, 140]))
    plt.savefig("../paper/figures/"+"kronecker_l2_norm_nonsparse.pdf")


def plot_ROC_curve(figure_name):
    """
    plot information in a pretty way
    """
    plt.clf()
    fig = plt.figure(2)

    fig, ax = plt.subplots()

    recall_sparse_200 = [.03, .16, .37, .4, .49]
    precision_sparse_200 = [.9, .76, .61, .6, .63]

    recall_lasso_200 = [.02, .11, .25, .43, .5, .54]
    precision_lasso_200 = [.77, .77, .66, .56, .55, .51]

    recall_sparse_50 = [.07, .13, .16, .58]
    precision_sparse_50 = [.56, .53, .49, .37]

    recall_lasso_50 = [.03, .18, .27, .82]
    precision_lasso_50 = [.6, .47, .44, .24]

    plt.xlabel("Recall")
    plt.ylabel("Precision")
    plt.grid(color="lightgrey")

    ax.plot(recall_lasso_200, precision_lasso_200, 'ko-',
            color="lightseagreen", label="Lasso-200 cascades")
    ax.plot(recall_sparse_200, precision_sparse_200, 'ko-',
            color="k", label="Our Method-200 cascades")
    ax.plot(recall_lasso_50, precision_lasso_50, 'ko-',
            color="orange", label="Lasso-50 cascades")
    ax.plot(recall_sparse_50, precision_sparse_50, 'ko-',
            color="cornflowerblue", label="Our Method-50 cascades")
    plt.legend(loc="upper right", fontsize=14)
    plt.savefig("../paper/figures/"+"ROC_curve.pdf")


def plot_p_init_watts_strogatz():
    plt.clf()
    fig = plt.figure(1)
    x = [.01, .05, .1, .15, .2]
    greedy = [.43, .29, .18, .1, .08]
    sparse_recov = [.7, .58, .48, .39, .31]
    max_likel = [.69, .56, .45, .37, .3]
    lasso = [.66, .55, .46, .38, .3]

    fig, ax = plt.subplots()

    plt.subplots_adjust(bottom=.2, top=.85)
    plt.xticks(ha="right", rotation=45)

    plt.axis((0, .21, .05, .8))
    plt.xlabel("Number of Cascades", fontsize=20)
    plt.ylabel("F1-score", fontsize=20)
    plt.grid(color="lightgrey")
    ax.plot(x, greedy, 'ko-', color="lightseagreen", label='Greedy')
    ax.plot(x, lasso, 'ko-', color="orange", label="Lasso")
    ax.plot(x, max_likel, 'ko-', color="cornflowerblue", label="MLE")
    ax.plot(x, sparse_recov, 'ko-', color="k", label="Our Method")
    plt.legend(loc="upper right", fontsize=16)
    ax.set_xticks(x)
    ax.set_yticklabels(tuple([.1, .2, .3, .4, .5, .6, .7, .8]), fontsize=20)
    ax.set_xticklabels(tuple(x), fontsize=20)
    plt.savefig("../paper/figures/"+"watts_strogatz_p_init.pdf")



if __name__=="__main__":
    if 1:
        theoreticalGuarantees(graph_name =
                "../datasets/powerlaw_200_30_point3.txt",
                            n_cascades=100, min_proba=.2, max_proba=.7,
                            p_init=.05, sparse_edges=False,
                            passed_function=convex_optimization.sparse_recovery,
                            lbda=.001)

    if 0:
        cascades_vs_measurements("../datasets/watts_strogatz_p_init.txt",
                n_cascades=1000, min_proba=.2, max_proba=.7)
    if 0:
        compute_graph("../datasets/watts_strogatz_300_30_point3.txt",
                    n_cascades=300, lbda=.013382, min_proba=.2, max_proba=.7,
                    passed_function=
                    #convex_optimization.sparse_recovery)
                    algorithms.greedy_prediction, p_init=.2)
                    #convex_optimization.sparse_recovery, p_init=.15)
    if 0:
        compute_graph("../datasets/powerlaw_200_30_point3.txt",
            n_cascades=200, lbda=.01, min_proba=.2, max_proba=.7,
            passed_function=
            #convex_optimization.sparse_recovery)
            #algorithms.greedy_prediction)
            convex_optimization.type_lasso)
    if 0:
        compute_graph("../datasets/barabasi_albert_300_30.txt",
            n_cascades=100, lbda=.002, min_proba=.2,
            max_proba=.7, passed_function=
            convex_optimization.sparse_recovery)
            #algorithms.greedy_prediction)
            #convex_optimization.type_lasso)
    if 0:
        compute_graph("../datasets/kronecker_graph_256_cross.txt",
            n_cascades=50, lbda=0., min_proba=.2, max_proba=.7,
            passed_function=
            convex_optimization.sparse_recovery,
            #convex_optimization.type_lasso,
            sparse_edges=True)