import argparse
import numpy as np

parser = argparse.ArgumentParser()

parser.add_argument('task_num', type=int, help='Číslo úlohy')
parser.add_argument('--numpy', action='store_true', help='Použít numpy')


def _predict(data, weights):
    output = []
    features_num = len(weights) - 1
    for dato in data:
        predict = 0.0
        for ind in range(features_num):
            predict += weights[ind] * dato[ind]
        # Bias
        predict += weights[-1]
        output.append(predict)
    return output


def _parse_input(read_target=False):
    data_num, _ = [int(num) for num in input().split()]
    weights = [float(feature) for feature in input().split()]

    data = []
    target = []
    for _ in range(data_num):
        line = [float(feature) for feature in input().split()]

        if read_target:
            # Features + Bias
            data.append(line[:-1] + [1])
            target.append(line[-1])
        else:
            # Features + Bias
            data.append(line + [1])

    if read_target:
        return data, target, weights
    else:
        return data, weights


def task_predict():
    data, weights = _parse_input(read_target=False)

    output = _predict(data, weights)
    for predict in output:
        print(predict)


def task_mse():
    data, target, weights = _parse_input(read_target=True)

    output = _predict(data, weights)
    mse = 0
    for ind in range(len(target)):
        mse += (output[ind] - target[ind]) ** 2
    mse /= len(target)
    print(mse)


def tak_optimal_weights():
    data, target, weights = _parse_input(read_target=True)

    output = _predict(data, weights)

    max_diff = 0
    for idx_weight in range(len(weights)):
        diff = 0
        for idx_data in range(len(data)):
            diff += (target[idx_data] - output[idx_data]) * data[idx_data][idx_weight]
        diff = abs(diff)
        max_diff = max(max_diff, diff)

    if max_diff > 1e-6:
        print("NE")
    else:
        print("ANO")


def _parse_input_numpy(read_target=False):
    data = list(_parse_input(read_target=read_target))
    for idx in range(len(data)):
        data[idx] = np.array(data[idx])
    return data


def task_predict_numpy():
    data, weights = _parse_input_numpy(read_target=False)

    output = data @ weights

    for predict in output:
        print(predict)


def task_mse_numpy():
    data, target, weights = _parse_input_numpy(read_target=True)

    output = data @ weights
    mse = np.mean((output - target) ** 2)
    print(mse)


def tak_optimal_weights_numpy():
    data, target, weights = _parse_input_numpy(read_target=True)

    output = data @ weights
    max_diff = np.max(np.abs(data.T @ (target - output)))

    if max_diff > 1e-6:
        print("NE")
    else:
        print("ANO")


if __name__ == "__main__":
    args = parser.parse_args()
    tasks: dict = {1: task_predict, 2: task_mse, 3: tak_optimal_weights}

    function = tasks[args.task_num]
    if args.numpy:
        # Zavolání funkce s příponou _numpy
        function = locals()[function.__name__ + '_numpy']
    function()