How to use forestci - 3 common examples

of Machine Learning Research vol. 15, pp. 1625-1651, 2014.
    """
    if inbag is None:
        inbag = calc_inbag_modified(X_train.shape[0], forest, is_ensemble)

    if not is_ensemble:
        pred = np.array([tree.predict(X_test) for tree in forest]).T
    else:
        pred = np.array([tree.predict(X_test) for tree in forest.model]).T
        pred = pred[0]
    pred_mean = np.mean(pred, 0)
    pred_centered = pred - pred_mean
    n_trees = forest.n_estimators
    V_IJ = fci._core_computation(X_train, X_test, inbag, pred_centered, n_trees,
                             memory_constrained, memory_limit)
    V_IJ_unbiased = fci._bias_correction(V_IJ, inbag, pred_centered, n_trees)

    # Correct for cases where resampling is done without replacement:
    if np.max(inbag) == 1:
        variance_inflation = 1 / (1 - np.mean(inbag)) ** 2
        V_IJ_unbiased *= variance_inflation

    if basic_IJ:
        return V_IJ

    if not calibrate:
        return V_IJ_unbiased

    if V_IJ_unbiased.shape[0] <= 20:
        print("No calibration with n_samples <= 20")
        return V_IJ_unbiased
    if calibrate:

.. [Wager2014] S. Wager, T. Hastie, B. Efron. "Confidence Intervals for
       Random Forests: The Jackknife and the Infinitesimal Jackknife", Journal
       of Machine Learning Research vol. 15, pp. 1625-1651, 2014.
    """
    if inbag is None:
        inbag = calc_inbag_modified(X_train.shape[0], forest, is_ensemble)

    if not is_ensemble:
        pred = np.array([tree.predict(X_test) for tree in forest]).T
    else:
        pred = np.array([tree.predict(X_test) for tree in forest.model]).T
        pred = pred[0]
    pred_mean = np.mean(pred, 0)
    pred_centered = pred - pred_mean
    n_trees = forest.n_estimators
    V_IJ = fci._core_computation(X_train, X_test, inbag, pred_centered, n_trees,
                             memory_constrained, memory_limit)
    V_IJ_unbiased = fci._bias_correction(V_IJ, inbag, pred_centered, n_trees)

    # Correct for cases where resampling is done without replacement:
    if np.max(inbag) == 1:
        variance_inflation = 1 / (1 - np.mean(inbag)) ** 2
        V_IJ_unbiased *= variance_inflation

    if basic_IJ:
        return V_IJ

    if not calibrate:
        return V_IJ_unbiased

    if V_IJ_unbiased.shape[0] <= 20:
        print("No calibration with n_samples <= 20")

#results_ss = fci.random_forest_error(new_forest, X_train, X_test,
        #                                 calibrate=False,
        #                                 memory_constrained=memory_constrained,
        #                                 memory_limit=memory_limit)
        results_ss = random_forest_error_modified(new_forest, is_ensemble, X_train, X_test,
                                         calibrate=False,
                                         memory_constrained=memory_constrained,
                                         memory_limit=memory_limit)
        # Use this second set of variance estimates
        # to estimate scale of Monte Carlo noise
        sigma2_ss = np.mean((results_ss - V_IJ_unbiased)**2)
        delta = n_sample / n_trees
        sigma2 = (delta**2 + (1 - delta)**2) / (2 * (1 - delta)**2) * sigma2_ss

        # Use Monte Carlo noise scale estimate for empirical Bayes calibration
        V_IJ_calibrated = fci.calibration.calibrateEB(V_IJ_unbiased, sigma2)

        return V_IJ_calibrated