zero_corr.py 文件源码

def plot_samples(samples, dist, noise, modelno, num_samples, timestamp):
    """Plot the observed samples and posterior samples side-by-side."""
    print 'Plotting samples %s %f' % (dist, noise)
    fig, ax = plt.subplots(nrows=1, ncols=2)
    fig.suptitle(
        '%s (noise %1.2f, sample %d)' % (dist, noise, modelno),
        size=16)
    # Plot the observed samples.
    T = simulate_dataset(dist, noise, num_samples)
    # ax[0].set_title('Observed Data')
    ax[0].text(
        .5, .95, 'Observed Data',
        horizontalalignment='center',
        transform=ax[0].transAxes)
    ax[0].set_xlabel('x1')
    ax[0].set_ylabel('x2')
    ax[0].scatter(T[:,0], T[:,1], color='k', alpha=.5)
    ax[0].set_xlim(simulator_limits[dist][0])
    ax[0].set_ylim(simulator_limits[dist][1])
    ax[0].grid()
    # Plot posterior distribution.
    # ax[1].set_title('CrossCat Posterior Samples')
    ax[1].text(
        .5, .95, 'CrossCat Posterior Samples',
        horizontalalignment='center',
        transform=ax[1].transAxes)
    ax[1].set_xlabel('x1')
    clusters = set(samples[:,2])
    colors = iter(matplotlib.cm.gist_rainbow(
        np.linspace(0, 1, len(clusters)+2)))
    for c in clusters:
        sc = samples[samples[:,2] == c][:,[0,1]]
        ax[1].scatter(sc[:,0], sc[:,1], alpha=.5, color=next(colors))
    ax[1].set_xlim(ax[0].get_xlim())
    ax[1].set_ylim(ax[0].get_ylim())
    ax[1].grid()
    # Save.
    # fig.set_tight_layout(True)
    fig.savefig(filename_samples_figure(dist, noise, modelno, timestamp))
    plt.close('all')