import numpy as np
import numpy.random
import scipy.special


class NeuralNetwork:

    # initialise the neural network
    def __init__(self, input_nodes_num, hidden_nodes_num, output_nodes_num, lr):
        # 初始化神經元個數，可以直接修改
        self.input_nodes = input_nodes_num
        self.hidden_nodes = hidden_nodes_num
        self.output_nodes 
 
= output_nodes_num
        self.learning_rate = lr

        # 初始化權重值，利用正態分佈函式進行隨機初始化，均值為0，方差為神經元個數開方
        self.w_input_hidden = numpy.random.normal(0.0, pow(self.hidden_nodes, -0.5),
                                                  (self.hidden_nodes, self.input_nodes))
        self.w_hidden_output 
 
= numpy.random.normal(0.0, pow(self.output_nodes, -0.5),
                                                   (self.output_nodes, self.hidden_nodes))
        # 初始化啟用函式，啟用函式選用Sigmoid函式，更加平滑，接近自然界的神經元行為模式
        # lambda定義了一個匿名函式
        self.activation_function = lambda x: scipy.special.expit(x)
        
 
pass

    # train the neural network
    def train(self, inputs_list, targets_list):
        # 將訓練集和測試集中的資料轉化為列向量
        inputs = np.array(inputs_list, ndmin=2).T
        targets = np.array(targets_list, ndmin=2).T
        # 隱藏層的輸入為訓練集與權重值的點乘，輸出為啟用函式的輸出
        hidden_inputs = np.dot(self.w_input_hidden, inputs)
        hidden_outputs = self.activation_function(hidden_inputs)
        # 輸出層的輸入為隱藏層的輸出，輸出為最終結果
        final_inputs = np.dot(self.w_hidden_output, hidden_outputs)
        final_outputs = self.activation_function(final_inputs)
        # 損失函式
        output_errors = targets - final_outputs
        # 隱藏層的誤差為權值矩陣的轉置與輸出誤差的點乘
        hidden_errors = np.dot(self.w_hidden_output.T, output_errors)
        # 對權值進行更新
        self.w_hidden_output += self.learning_rate * np.dot((output_errors *
                                                             final_outputs * (1.0 - final_outputs)),
                                                            np.transpose(hidden_outputs))

        self.w_input_hidden += self.learning_rate * np.dot((hidden_errors *
                                                            hidden_outputs * (1.0 - hidden_outputs)),
                                                           np.transpose(inputs))

        pass

    # query the neural network
    def query(self, inputs_list):
        # 轉置將行向量轉成列向量，將每組資料更好的分隔開來，方便後續矩陣點乘操作
        inputs = np.array(inputs_list, ndmin=2).T
        # 加權求和後經過sigmoid函式得到隱藏層輸出
        hidden_inputs = np.dot(self.w_input_hidden, inputs)
        hidden_outputs = self.activation_function(hidden_inputs)
        # 加權求和後經過sigmoid函式得到最終輸出
        final_inputs = np.dot(self.w_hidden_output, hidden_outputs)
        final_outputs = self.activation_function(final_inputs)
        # 得到輸出資料列
        return final_outputs


# 初始化各層神經元個數，期中輸入神經元個數取決於讀入的因變數，而輸出神經元個數取決於分類的可能性個數
input_nodes = 784
hidden_nodes = 500
output_nodes = 10
# 學習率，每次調整步幅大小
learning_rate = 0.2

n = NeuralNetwork(input_nodes, hidden_nodes, output_nodes, learning_rate)
# 獲取訓練集資訊
training_data_file = open('data/mnist_train.csv', 'r')
training_data_list = training_data_file.readlines()
training_data_file.close()

for record in training_data_list:
    all_values = record.split(',')

    inputs = (numpy.asfarray(all_values[1:]) / 255.0 * 0.99) + 0.01

    targets = numpy.zeros(output_nodes) + 0.01
    targets[int(all_values[0])] = 0.99
    n.train(inputs, targets)
    pass
print('train successful!')
test_file = open('data/mnist_test.csv', 'r')
test_list = test_file.readlines()
test_file.close()
m = np.size(test_list)
j = 0.0
for record in test_list:
    test_values = record.split(',')
    np.asfarray(test_values)
    results = n.query(np.asfarray(test_values[1:]))
    if results[int(test_values[0])] == max(results):
        j += 1
    pass

print("正確率為；" + str(j/m))