---

import numpy as np


# функция активации ReLU
def relu(x):
    return max(0, x)


# функция активации tanh
def tanh(x):
    return np.tanh(x)


# класс нейрона
class Neuron:
    def __init__(self, weights, bias, activation):
        self.weights = weights
        self.bias = bias
        self.activation = activation

    def feedforward(self, inputs):
        total = np.dot(self.weights, inputs) + self.bias
        return self.activation(total)


# нейросеть с функцией активации ReLU
class NeuralNetworkReLU:
    def __init__(self):
        weights = np.array([0.5, 0.5, 0.5])
        bias = 0

        self.h1 = Neuron(weights, bias, relu)
        self.h2 = Neuron(weights, bias, relu)
        self.h3 = Neuron(weights, bias, relu)

        self.o1 = Neuron(weights, bias, relu)

    def feedforward(self, x):
        out_h1 = self.h1.feedforward(x)
        out_h2 = self.h2.feedforward(x)
        out_h3 = self.h3.feedforward(x)

        out_hidden = np.array([out_h1, out_h2, out_h3])

        out_o1 = self.o1.feedforward(out_hidden)

        return out_o1


# нейросеть с функцией активации tanh
class NeuralNetworkTanh:
    def __init__(self):
        weights = np.array([0.5, 0.5, 0.5])
        bias = 0

        self.h1 = Neuron(weights, bias, tanh)
        self.h2 = Neuron(weights, bias, tanh)
        self.h3 = Neuron(weights, bias, tanh)

        self.o1 = Neuron(weights, bias, tanh)

    def feedforward(self, x):
        out_h1 = self.h1.feedforward(x)
        out_h2 = self.h2.feedforward(x)
        out_h3 = self.h3.feedforward(x)

        out_hidden = np.array([out_h1, out_h2, out_h3])

        out_o1 = self.o1.feedforward(out_hidden)

        return out_o1


# входные данные
x = np.array([2, 3, 4])

# проверка ReLU-сети
network_relu = NeuralNetworkReLU()
print("Результат нейросети с ReLU:")
print(network_relu.feedforward(x))

# проверка tanh-сети
network_tanh = NeuralNetworkTanh()
print("Результат нейросети с tanh:")
print(network_tanh.feedforward(x))