def plot_roc(probas, y_true):
    plt.figure(figsize=(15,5))
    mean_tpr = 0.0
    mean_fpr = np.linspace(0, 1, 100)
    all_tpr = []

    classes = np.unique(y_true)
    perclass_mean_tpr = 0.0
    roc_auc = 0
    for j in classes:
        fpr, tpr, thresholds = mt.roc_curve(y_true, probas[:, j], pos_label=j)
        perclass_mean_tpr += interp(mean_fpr, fpr, tpr)
        perclass_mean_tpr[0] = 0.0
        roc_auc += mt.auc(fpr, tpr)
        plt.plot(fpr,tpr,'--',lw=.5,label='Class ROC for ensemble, AUC=%0.4f'
                     %(mt.auc(fpr, tpr)) )

    perclass_mean_tpr /= len(classes)
    roc_auc /= len(classes)
    mean_tpr += perclass_mean_tpr

    plt.plot(mean_fpr,perclass_mean_tpr,'-',lw=2,label='Mean Class ROC for ensemble, AUC=%0.4f'
                     %(roc_auc))
    plt.legend(loc='best')
    plt.xlabel('false positive rate')
    plt.ylabel('true positive rate')
    plt.title('ROC Curve')